{ "generatedAt": "2026-09-01T03:04:52+00:00", "conventions": { "capabilityBenchmark": "GPQA diamond", "complexTokens": 7000, "chatTokens": 300, "grid": "world average", "datacenterPUE": 1.2, "calibrationFactor": 0.53518, "calibrationBasis": "46 samples vs ML.ENERGY on B200, batch 32-128", "reasoningDisclosureReviewed": "2026-09-01" }, "counts": { "total": 178, "withEnergy": 134, "modeled": 77, "day1_estimate": 57, "corroboratedByLab": 11, "unmatched": 44, "reasoningHidden": 74, "reasoningShown": 75, "reasoningUndisclosed": 29 }, "models": [ { "id": "gemini37flash", "name": "Gemini 3.7 Flash", "org": "Google DeepMind", "releaseDate": "2026-08-13", "openWeights": false, "capability": 94.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.47, "FrontierMath-Tier-4-v2-Private": 0.3659, "FrontierMath-Tiers-1-3-v2-Private": 0.7158, "GPQA diamond": 0.9482, "Mystery Game Puzzles": 0.37, "OTIS Mock AIME 2024-2025": 0.9722, "SimpleQA Verified": 0.692 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "gpt54pro", "name": "GPT-5.4 Pro", "org": "OpenAI", "releaseDate": "2026-03-05", "openWeights": false, "capability": 94.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.586, "FrontierMath-2025-02-28-Private": 0.5, "FrontierMath-Tier-4-2025-07-01-Private": 0.375, "FrontierMath-Tier-4-v2-Private": 0.5854, "FrontierMath-Tiers-1-3-v2-Private": 0.8246, "GPQA diamond": 0.946, "SimpleQA Verified": 0.463 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 42.88592, "energyChatRange": [ 25.192501, 60.57934 ], "energyComplex": 1000.671477, "energyComplexRange": [ 587.825014, 1413.517939 ], "activeParamsB": [ 540, 1800 ], "totalParamsB": 5400.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 4.0, "isParetoFrontier": true }, { "id": "gemini31pro", "name": "Gemini 3.1 Pro", "org": "Google DeepMind", "releaseDate": "2026-02-19", "openWeights": false, "capability": 94.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.55, "EBR-bench": 0.1429, "FrontierMath-2025-02-28-Private": 0.369, "FrontierMath-2025-02-28-Public": 0.8889, "FrontierMath-Tier-4-2025-07-01-Private": 0.167, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tier-4-v2-Private": 0.2683, "FrontierMath-Tiers-1-3-v2-Private": 0.5965, "GPQA diamond": 0.941, "MirrorCode": 0.0889, "Mystery Game Puzzles": 0.34, "OTIS Mock AIME 2024-2025": 0.956, "SWE-Bench verified": 0.7562, "SimpleQA Verified": 0.735 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 2.294805, "energyChatRange": [ 1.452262, 3.137349 ], "energyComplex": 53.545461, "energyComplexRange": [ 33.886106, 73.204817 ], "activeParamsB": [ 120, 360 ], "totalParamsB": 1200.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 73.0, "isParetoFrontier": true }, { "id": "gemini36flash", "name": "Gemini 3.6 Flash", "org": "Google DeepMind", "releaseDate": "2026-07-21", "openWeights": false, "capability": 94.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.4, "FrontierMath-Tier-4-v2-Private": 0.2195, "FrontierMath-Tiers-1-3-v2-Private": 0.5895, "GPQA diamond": 0.9413, "Mystery Game Puzzles": 0.3, "OTIS Mock AIME 2024-2025": 0.9417, "SimpleQA Verified": 0.662 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "gpt55", "name": "GPT-5.5", "org": "OpenAI", "releaseDate": "2026-04-23", "openWeights": false, "capability": 94.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.54, "EBR-bench": 0.3429, "FrontierMath-2025-02-28-Private": 0.517, "FrontierMath-Tier-4-2025-07-01-Private": 0.354, "FrontierMath-Tier-4-v2-Private": 0.725, "FrontierMath-Tiers-1-3-v2-Private": 0.8526, "GPQA diamond": 0.94, "MirrorCode": 0.1, "Mystery Game Puzzles": 0.56, "OEIS Open": 0.2622, "OEIS Open Lite": 0.36, "OTIS Mock AIME 2024-2025": 1.0, "SWE-Bench verified": 0.8058, "SimpleQA Verified": 0.63 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 1.033794, "energyChatRange": [ 0.665181, 1.402407 ], "energyComplex": 24.121868, "energyComplexRange": [ 15.5209, 32.722836 ], "activeParamsB": [ 90, 300 ], "totalParamsB": 900.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 40.4, "isParetoFrontier": true }, { "id": "grok46", "name": "Grok 4.6", "org": "xAI", "releaseDate": "2026-08-12", "openWeights": false, "capability": 94.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.4, "EBR-bench": 0.3048, "FrontierMath-Tier-4-v2-Private": 0.3171, "FrontierMath-Tiers-1-3-v2-Private": 0.6596, "GPQA diamond": 0.94, "Mystery Game Puzzles": 0.34, "OTIS Mock AIME 2024-2025": 0.9778, "SimpleQA Verified": 0.489 }, "reasoning": { "status": "undisclosed", "note": "xAI publishes no position on whether reasoning text is returned, and no parameter counts either. Nothing here is assumed.", "source": null }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "claudeopus5", "name": "Claude Opus 5", "org": "Anthropic", "releaseDate": "2026-07-24", "openWeights": false, "capability": 93.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.42, "EBR-bench": 0.5, "FrontierMath-Tier-4-v2-Private": 0.7317, "FrontierMath-Tiers-1-3-v2-Private": 0.8561, "GPQA diamond": 0.9388, "Mystery Game Puzzles": 0.59, "OTIS Mock AIME 2024-2025": 0.9889, "SimpleQA Verified": 0.599 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "gpt55pro", "name": "GPT-5.5 Pro", "org": "OpenAI", "releaseDate": "2026-04-23", "openWeights": false, "capability": 93.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.64, "FrontierMath-2025-02-28-Private": 0.524, "FrontierMath-Tier-4-2025-07-01-Private": 0.396, "FrontierMath-Tier-4-v2-Private": 0.7805, "FrontierMath-Tiers-1-3-v2-Private": 0.8772, "GPQA diamond": 0.9392, "OTIS Mock AIME 2024-2025": 1.0 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 37.088126, "energyChatRange": [ 19.394706, 54.781546 ], "energyComplex": 865.389605, "energyComplexRange": [ 452.543142, 1278.236067 ], "activeParamsB": [ 540, 1800 ], "totalParamsB": 5400.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 14.4, "isParetoFrontier": false }, { "id": "gpt56sol", "name": "GPT-5.6 Sol", "org": "OpenAI", "releaseDate": "2026-07-09", "openWeights": false, "capability": 93.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.55, "EBR-bench": 0.3905, "FrontierMath-Tier-4-v2-Private": 0.8293, "FrontierMath-Tiers-1-3-v2-Private": 0.8912, "GPQA diamond": 0.935, "MirrorCode": 0.2, "Mystery Game Puzzles": 0.58, "OEIS Open Lite": 0.43, "OTIS Mock AIME 2024-2025": 1.0, "SimpleQA Verified": 0.697 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "grok45", "name": "Grok 4.5", "org": "xAI", "releaseDate": "2026-07-08", "openWeights": false, "capability": 93.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.36, "FrontierMath-Tier-4-v2-Private": 0.2439, "FrontierMath-Tiers-1-3-v2-Private": 0.5719, "GPQA diamond": 0.9343, "OTIS Mock AIME 2024-2025": 0.9778, "SimpleQA Verified": 0.483 }, "reasoning": { "status": "undisclosed", "note": "xAI publishes no position on whether reasoning text is returned, and no parameter counts either. Nothing here is assumed.", "source": null }, "energyChat": 6.191215, "energyChatRange": [ 1.083294, 11.299137 ], "energyComplex": 144.461693, "energyComplexRange": [ 25.276851, 263.646535 ], "activeParamsB": [ 45.0, 1500.0 ], "totalParamsB": 1500.0, "sizeSource": "epoch_arch_unknown", "architectureKnown": false, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "gpt54", "name": "GPT-5.4", "org": "OpenAI", "releaseDate": "2026-03-05", "openWeights": false, "capability": 93.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.44, "EBR-bench": 0.254, "FrontierMath-2025-02-28-Private": 0.476, "FrontierMath-2025-02-28-Public": 0.8, "FrontierMath-Tier-4-2025-07-01-Private": 0.271, "FrontierMath-Tier-4-2025-07-01-Public": 0.5, "FrontierMath-Tier-4-v2-Private": 0.49, "FrontierMath-Tiers-1-3-v2-Private": 0.786, "GPQA diamond": 0.933, "MirrorCode": 0.1556, "Mystery Game Puzzles": 0.37, "OTIS Mock AIME 2024-2025": 0.953, "SWE-Bench verified": 0.7686, "SimpleQA Verified": 0.451 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.340187, "energyChatRange": [ 0.248034, 0.432341 ], "energyComplex": 7.937704, "energyComplexRange": [ 5.787462, 10.087946 ], "activeParamsB": [ 45, 150 ], "totalParamsB": 450.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 45.5, "isParetoFrontier": true }, { "id": "gpt56terra", "name": "GPT-5.6 Terra", "org": "OpenAI", "releaseDate": "2026-07-09", "openWeights": false, "capability": 93.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.54, "FrontierMath-Tier-4-v2-Private": 0.7073, "FrontierMath-Tiers-1-3-v2-Private": 0.8596, "GPQA diamond": 0.9331, "Mystery Game Puzzles": 0.35, "OTIS Mock AIME 2024-2025": 0.9972, "SimpleQA Verified": 0.432 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "kimik3", "name": "Kimi K3", "org": "Moonshot", "releaseDate": "2026-07-16", "openWeights": true, "capability": 93.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.39, "FrontierMath-Tier-4-v2-Private": 0.3902, "FrontierMath-Tiers-1-3-v2-Private": 0.7218, "GPQA diamond": 0.9312, "Mystery Game Puzzles": 0.26, "OTIS Mock AIME 2024-2025": 0.9722, "SimpleQA Verified": 0.506 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 2.995089, "energyChatRange": [ 2.995089, 2.995089 ], "energyComplex": 69.885402, "energyComplexRange": [ 69.885402, 69.885402 ], "activeParamsB": 104.0, "totalParamsB": 2800.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gemini35flash", "name": "Gemini 3.5 Flash", "org": "Google", "releaseDate": "2026-05-19", "openWeights": false, "capability": 92.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.5, "EBR-bench": 0.0476, "FrontierMath-2025-02-28-Private": 0.3897, "FrontierMath-2025-02-28-Public": 0.8, "FrontierMath-Tier-4-2025-07-01-Private": 0.1458, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tier-4-v2-Private": 0.2683, "FrontierMath-Tiers-1-3-v2-Private": 0.6281, "GPQA diamond": 0.928, "Mystery Game Puzzles": 0.32, "OEIS Open": 0.2215, "OEIS Open Lite": 0.29, "OTIS Mock AIME 2024-2025": 0.9556, "SWE-Bench verified": 0.7934, "SimpleQA Verified": 0.662 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 0.859857, "energyChatRange": [ 0.596562, 1.123152 ], "energyComplex": 20.06334, "energyComplexRange": [ 13.919792, 26.206889 ], "activeParamsB": [ 90, 240 ], "totalParamsB": 900.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 130.6, "isParetoFrontier": false }, { "id": "qwen38max", "name": "Qwen 3.8 Max", "org": "Alibaba", "releaseDate": "2026-08-02", "openWeights": false, "capability": 92.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.29, "FrontierMath-Tier-4-v2-Private": 0.4634, "FrontierMath-Tiers-1-3-v2-Private": 0.7474, "GPQA diamond": 0.9268, "Mystery Game Puzzles": 0.38, "OTIS Mock AIME 2024-2025": 0.9944, "SimpleQA Verified": 0.458 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": 2.868707, "energyChatRange": [ 2.868707, 2.868707 ], "energyComplex": 66.936498, "energyComplexRange": [ 66.936498, 66.936498 ], "activeParamsB": 95.0, "totalParamsB": 2400.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gemini3pro", "name": "Gemini 3 Pro", "org": "Google DeepMind", "releaseDate": "2025-11-18", "openWeights": false, "capability": 92.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.31, "FrontierMath-2025-02-28-Private": 0.376, "FrontierMath-2025-02-28-Public": 0.8, "FrontierMath-Tier-4-2025-07-01-Private": 0.1875, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "GPQA diamond": 0.9261, "OTIS Mock AIME 2024-2025": 0.9139, "SWE-Bench verified": 0.7293 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 2.294805, "energyChatRange": [ 1.452262, 3.137349 ], "energyComplex": 53.545461, "energyComplexRange": [ 33.886106, 73.204817 ], "activeParamsB": [ 120, 360 ], "totalParamsB": 1200.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 73.0, "isParetoFrontier": false }, { "id": "glm52", "name": "GLM-5.2", "org": "Z.ai (Zhipu AI)", "releaseDate": "2026-06-16", "openWeights": true, "capability": 91.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.21, "EBR-bench": 0.0952, "FrontierMath-Tier-4-v2-Private": 0.2927, "FrontierMath-Tiers-1-3-v2-Private": 0.5921, "GPQA diamond": 0.9186, "Mystery Game Puzzles": 0.18, "OTIS Mock AIME 2024-2025": 0.8639, "SWE-Bench verified": 0.787, "SimpleQA Verified": 0.342 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.524094, "energyChatRange": [ 0.524094, 0.524094 ], "energyComplex": 12.228856, "energyComplexRange": [ 12.228856, 12.228856 ], "activeParamsB": 40.0, "totalParamsB": 744.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "deepseekv4pro0813", "name": "DeepSeek V4 Pro 0813", "org": "DeepSeek", "releaseDate": "2026-08-13", "openWeights": true, "capability": 91.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.47, "FrontierMath-Tier-4-v2-Private": 0.2683, "FrontierMath-Tiers-1-3-v2-Private": 0.6456, "GPQA diamond": 0.9167, "Mystery Game Puzzles": 0.43, "OTIS Mock AIME 2024-2025": 0.9861, "SimpleQA Verified": 0.5291 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 1.111378, "energyChatRange": [ 1.111378, 1.111378 ], "energyComplex": 25.932163, "energyComplexRange": [ 25.932163, 25.932163 ], "activeParamsB": 49.0, "totalParamsB": 1600.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gpt56luna", "name": "GPT-5.6 Luna", "org": "OpenAI", "releaseDate": "2026-07-09", "openWeights": false, "capability": 91.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.4, "FrontierMath-Tier-4-v2-Private": 0.6098, "FrontierMath-Tiers-1-3-v2-Private": 0.8211, "GPQA diamond": 0.916, "Mystery Game Puzzles": 0.21, "OTIS Mock AIME 2024-2025": 0.9833, "SimpleQA Verified": 0.41 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "claudeopus48", "name": "Claude Opus 4.8", "org": "Anthropic", "releaseDate": "2026-05-28", "openWeights": false, "capability": 91.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.34, "EBR-bench": 0.2857, "FrontierMath-2025-02-28-Private": 0.4724, "FrontierMath-Tier-4-2025-07-01-Private": 0.3125, "FrontierMath-Tier-4-v2-Private": 0.561, "FrontierMath-Tiers-1-3-v2-Private": 0.8, "GPQA diamond": 0.9104, "Mystery Game Puzzles": 0.36, "OEIS Open": 0.2988, "OEIS Open Lite": 0.39, "OTIS Mock AIME 2024-2025": 0.9833, "SimpleQA Verified": 0.53 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": 0.787744, "energyChatRange": [ 0.554289, 1.021199 ], "energyComplex": 18.380694, "energyComplexRange": [ 12.933414, 23.827974 ], "activeParamsB": [ 67, 200 ], "totalParamsB": 670.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 58.0, "isParetoFrontier": false }, { "id": "deepseekv4flash0731", "name": "DeepSeek V4 Flash 0731", "org": "DeepSeek", "releaseDate": "2026-07-31", "openWeights": true, "capability": 91.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.33, "FrontierMath-Tier-4-v2-Private": 0.2439, "FrontierMath-Tiers-1-3-v2-Private": 0.5754, "GPQA diamond": 0.9104, "Mystery Game Puzzles": 0.34, "OTIS Mock AIME 2024-2025": 0.9444, "SimpleQA Verified": 0.3363 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.69034, "energyChatRange": [ 0.69034, 0.69034 ], "energyComplex": 16.107933, "energyComplexRange": [ 16.107933, 16.107933 ], "activeParamsB": 284.0, "totalParamsB": 284.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "glm53", "name": "GLM-5.3", "org": "Z.ai (Zhipu AI)", "releaseDate": "2026-08-14", "openWeights": false, "capability": 90.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.21, "FrontierMath-Tier-4-v2-Private": 0.2927, "FrontierMath-Tiers-1-3-v2-Private": 0.6877, "GPQA diamond": 0.9091, "Mystery Game Puzzles": 0.33, "OTIS Mock AIME 2024-2025": 0.9111, "SimpleQA Verified": 0.41 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": 2.995556, "energyChatRange": [ 2.995556, 2.995556 ], "energyComplex": 69.896298, "energyComplexRange": [ 69.896298, 69.896298 ], "activeParamsB": 744.0, "totalParamsB": 744.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "minimaxm3", "name": "MiniMax-M3", "org": "MiniMax", "releaseDate": "2026-06-01", "openWeights": true, "capability": 90.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.14, "GPQA diamond": 0.9091, "Mystery Game Puzzles": 0.08, "OTIS Mock AIME 2024-2025": 0.7111 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.943103, "energyChatRange": [ 0.943103, 0.943103 ], "energyComplex": 22.00574, "energyComplexRange": [ 22.00574, 22.00574 ], "activeParamsB": 428.0, "totalParamsB": 428.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "qwen37max", "name": "Qwen3.7-Max", "org": "Alibaba", "releaseDate": "2026-05-19", "openWeights": false, "capability": 90.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.19, "EBR-bench": 0.0952, "FrontierMath-Tier-4-v2-Private": 0.3415, "FrontierMath-Tiers-1-3-v2-Private": 0.6456, "GPQA diamond": 0.9091, "Mystery Game Puzzles": 0.32, "OTIS Mock AIME 2024-2025": 0.9556, "SWE-Bench verified": 0.7727, "SimpleQA Verified": 0.5577 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "kimik26", "name": "Kimi K2.6", "org": "Moonshot", "releaseDate": "2026-04-20", "openWeights": true, "capability": 90.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.26, "EBR-bench": 0.0238, "FrontierMath-2025-02-28-Private": 0.3897, "FrontierMath-2025-02-28-Public": 0.9, "FrontierMath-Tier-4-2025-07-01-Private": 0.1458, "FrontierMath-Tier-4-v2-Private": 0.2564, "FrontierMath-Tiers-1-3-v2-Private": 0.5719, "GPQA diamond": 0.9078, "Mystery Game Puzzles": 0.18, "OTIS Mock AIME 2024-2025": 0.9611, "SWE-Bench verified": 0.7665, "SimpleQA Verified": 0.349 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 2.263947, "energyChatRange": [ 0.493201, 4.034693 ], "energyComplex": 52.825424, "energyComplexRange": [ 11.508013, 94.142836 ], "activeParamsB": [ 31.2, 1040.0 ], "totalParamsB": 1040.0, "sizeSource": "epoch_arch_unknown", "architectureKnown": false, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "claudesonnet5", "name": "Claude Sonnet 5", "org": "Anthropic", "releaseDate": "2026-06-30", "openWeights": false, "capability": 90.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.16, "FrontierMath-Tier-4-v2-Private": 0.2927, "FrontierMath-Tiers-1-3-v2-Private": 0.6561, "GPQA diamond": 0.9053, "Mystery Game Puzzles": 0.35, "OTIS Mock AIME 2024-2025": 0.9472, "SimpleQA Verified": 0.337 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "claudeopus47", "name": "Claude Opus 4.7", "org": "Anthropic", "releaseDate": "2026-04-16", "openWeights": false, "capability": 90.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.3, "EBR-bench": 0.1905, "FrontierMath-2025-02-28-Private": 0.4379, "FrontierMath-2025-02-28-Public": 0.9, "FrontierMath-Tier-4-2025-07-01-Private": 0.2292, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tier-4-v2-Private": 0.3171, "FrontierMath-Tiers-1-3-v2-Private": 0.7018, "GPQA diamond": 0.9015, "MirrorCode": 0.3111, "Mystery Game Puzzles": 0.28, "OTIS Mock AIME 2024-2025": 0.978, "SWE-Bench verified": 0.8347, "SimpleQA Verified": 0.517 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": 0.80322, "energyChatRange": [ 0.569765, 1.036675 ], "energyComplex": 18.741803, "energyComplexRange": [ 13.294523, 24.189083 ], "activeParamsB": [ 67, 200 ], "totalParamsB": 670.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 47.4, "isParetoFrontier": false }, { "id": "glm53flash", "name": "GLM-5.3-Flash", "org": "Z.ai (Zhipu AI)", "releaseDate": "2026-08-20", "openWeights": true, "capability": 90.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.14, "FrontierMath-Tier-4-v2-Private": 0.1707, "FrontierMath-Tiers-1-3-v2-Private": 0.5579, "GPQA diamond": 0.9015, "Mystery Game Puzzles": 0.08, "OTIS Mock AIME 2024-2025": 0.9389 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "glm51", "name": "GLM-5.1", "org": "Z.ai (Zhipu AI)", "releaseDate": "2026-04-07", "openWeights": true, "capability": 89.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.19, "FrontierMath-2025-02-28-Private": 0.3345, "FrontierMath-2025-02-28-Public": 0.8, "FrontierMath-Tier-4-2025-07-01-Private": 0.125, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tiers-1-3-v2-Private": 0.2491, "GPQA diamond": 0.899, "OTIS Mock AIME 2024-2025": 0.9333, "SWE-Bench verified": 0.7417, "SimpleQA Verified": 0.34 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 1.746871, "energyChatRange": [ 0.46308, 3.030662 ], "energyComplex": 40.760314, "energyComplexRange": [ 10.805191, 70.715438 ], "activeParamsB": [ 22.619999999999997, 754.0 ], "totalParamsB": 754.0, "sizeSource": "epoch_arch_unknown", "architectureKnown": false, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "musespark", "name": "Muse Spark", "org": "Meta AI", "releaseDate": "2026-04-08", "openWeights": false, "capability": 89.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.39, "FrontierMath-Tier-4-2025-07-01-Private": 0.146, "GPQA diamond": 0.898, "OTIS Mock AIME 2024-2025": 0.889 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "deepseekv4pro", "name": "DeepSeek-V4-Pro", "org": "DeepSeek", "releaseDate": "2026-04-24", "openWeights": true, "capability": 89.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.2, "FrontierMath-Tier-4-v2-Private": 0.0244, "FrontierMath-Tiers-1-3-v2-Private": 0.4526, "GPQA diamond": 0.8965, "Mystery Game Puzzles": 0.17, "OTIS Mock AIME 2024-2025": 0.9667, "SWE-Bench verified": 0.7764, "SimpleQA Verified": 0.4699 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 1.111378, "energyChatRange": [ 1.111378, 1.111378 ], "energyComplex": 25.932163, "energyComplexRange": [ 25.932163, 25.932163 ], "activeParamsB": 49.0, "totalParamsB": 1600.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "grok420", "name": "Grok 4.20", "org": "xAI", "releaseDate": "2026-02-17", "openWeights": false, "capability": 89.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.24, "FrontierMath-Tier-4-v2-Private": 0.1707, "FrontierMath-Tiers-1-3-v2-Private": 0.4491, "GPQA diamond": 0.8933, "OTIS Mock AIME 2024-2025": 0.9222, "SimpleQA Verified": 0.302 }, "reasoning": { "status": "undisclosed", "note": "xAI publishes no position on whether reasoning text is returned, and no parameter counts either. Nothing here is assumed.", "source": null }, "energyChat": 0.643825, "energyChatRange": [ 0.218164, 1.069485 ], "energyComplex": 15.022573, "energyComplexRange": [ 5.090503, 24.954643 ], "activeParamsB": [ 15.0, 500.0 ], "totalParamsB": 500.0, "sizeSource": "epoch_arch_unknown", "architectureKnown": false, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "claudeopus46", "name": "Claude Opus 4.6", "org": "Anthropic", "releaseDate": "2026-02-05", "openWeights": false, "capability": 88.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.17, "EBR-bench": 0.127, "FrontierMath-2025-02-28-Private": 0.3828, "FrontierMath-2025-02-28-Public": 0.8, "FrontierMath-Tier-4-2025-07-01-Private": 0.1458, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tier-4-v2-Private": 0.2683, "FrontierMath-Tiers-1-3-v2-Private": 0.6596, "GPQA diamond": 0.8876, "Mystery Game Puzzles": 0.25, "OTIS Mock AIME 2024-2025": 0.9306, "SWE-Bench verified": 0.7562, "SimpleQA Verified": 0.47 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": 0.827022, "energyChatRange": [ 0.593567, 1.060477 ], "energyComplex": 19.297185, "energyComplexRange": [ 13.849905, 24.744465 ], "activeParamsB": [ 67, 200 ], "totalParamsB": 670.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 37.0, "isParetoFrontier": false }, { "id": "grok43beta", "name": "Grok 4.3 Beta", "org": "xAI", "releaseDate": "2026-04-17", "openWeights": false, "capability": 88.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.25, "FrontierMath-Tier-4-v2-Private": 0.1463, "FrontierMath-Tiers-1-3-v2-Private": 0.4281, "GPQA diamond": 0.8883, "OTIS Mock AIME 2024-2025": 0.9333, "SimpleQA Verified": 0.332 }, "reasoning": { "status": "undisclosed", "note": "xAI publishes no position on whether reasoning text is returned, and no parameter counts either. Nothing here is assumed.", "source": null }, "energyChat": 0.643825, "energyChatRange": [ 0.218164, 1.069485 ], "energyComplex": 15.022573, "energyComplexRange": [ 5.090503, 24.954643 ], "activeParamsB": [ 15.0, 500.0 ], "totalParamsB": 500.0, "sizeSource": "epoch_arch_unknown", "architectureKnown": false, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "inklingsmall", "name": "Inkling-Small", "org": "Thinking Machines", "releaseDate": "2026-07-15", "openWeights": true, "capability": 88.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.18, "FrontierMath-Tier-4-v2-Private": 0.1707, "FrontierMath-Tiers-1-3-v2-Private": 0.4632, "GPQA diamond": 0.8851, "Mystery Game Puzzles": 0.06, "OTIS Mock AIME 2024-2025": 0.9, "SimpleQA Verified": 0.191 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.212899, "energyChatRange": [ 0.212899, 0.212899 ], "energyComplex": 4.967632, "energyComplexRange": [ 4.967632, 4.967632 ], "activeParamsB": 12.0, "totalParamsB": 276.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": true }, { "id": "qwen36plus", "name": "Qwen 3.6 Plus", "org": "Alibaba", "releaseDate": "2026-03-31", "openWeights": false, "capability": 88.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.17, "FrontierMath-2025-02-28-Private": 0.2621, "FrontierMath-2025-02-28-Public": 0.5, "FrontierMath-Tier-4-2025-07-01-Private": 0.0833, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tiers-1-3-v2-Private": 0.3228, "GPQA diamond": 0.8838, "Mystery Game Puzzles": 0.12, "OTIS Mock AIME 2024-2025": 0.9333, "SWE-Bench verified": 0.5785, "SimpleQA Verified": 0.4414 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "inkling", "name": "Inkling", "org": "Thinking Machines", "releaseDate": "2026-07-15", "openWeights": true, "capability": 88.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.21, "FrontierMath-Tier-4-v2-Private": 0.0488, "FrontierMath-Tiers-1-3-v2-Private": 0.3333, "GPQA diamond": 0.8826, "OTIS Mock AIME 2024-2025": 0.8889, "SimpleQA Verified": 0.403 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 3.806504, "energyChatRange": [ 3.806504, 3.806504 ], "energyComplex": 88.818428, "energyComplexRange": [ 88.818428, 88.818428 ], "activeParamsB": 975.0, "totalParamsB": 975.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "kimik27code", "name": "Kimi K2.7 Code", "org": "Moonshot", "releaseDate": "2026-06-12", "openWeights": true, "capability": 87.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.21, "FrontierMath-Tier-4-v2-Private": 0.122, "FrontierMath-Tiers-1-3-v2-Private": 0.5404, "GPQA diamond": 0.8788, "OTIS Mock AIME 2024-2025": 0.9556, "SimpleQA Verified": 0.365 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 2.263947, "energyChatRange": [ 0.493201, 4.034693 ], "energyComplex": 52.825424, "energyComplexRange": [ 11.508013, 94.142836 ], "activeParamsB": [ 31.2, 1040.0 ], "totalParamsB": 1040.0, "sizeSource": "epoch_arch_unknown", "architectureKnown": false, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "glm5", "name": "GLM-5", "org": "Z.ai (Zhipu AI)", "releaseDate": "2026-02-11", "openWeights": true, "capability": 87.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.1, "FrontierMath-2025-02-28-Private": 0.1643, "FrontierMath-2025-02-28-Public": 0.0208, "FrontierMath-Tier-4-2025-07-01-Private": 0.021, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "GPQA diamond": 0.8782, "OTIS Mock AIME 2024-2025": 0.8, "SWE-Bench verified": 0.7208 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.644858, "energyChatRange": [ 0.488145, 0.801572 ], "energyComplex": 15.046697, "energyComplexRange": [ 11.390057, 18.703337 ], "activeParamsB": [ 29.76, 119.04 ], "totalParamsB": 744.0, "sizeSource": "epoch_moe_estimated", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "gpt51", "name": "GPT-5.1", "org": "OpenAI", "releaseDate": "2025-11-13", "openWeights": false, "capability": 87.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.32, "FrontierMath-2025-02-28-Private": 0.3103, "FrontierMath-Tier-4-2025-07-01-Private": 0.125, "GPQA diamond": 0.8763, "Mystery Game Puzzles": 0.15, "OTIS Mock AIME 2024-2025": 0.8861, "SWE-Bench verified": 0.6591, "SimpleQA Verified": 0.48 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.263373, "energyChatRange": [ 0.210714, 0.316032 ], "energyComplex": 6.145368, "energyComplexRange": [ 4.916658, 7.374078 ], "activeParamsB": [ 30, 90 ], "totalParamsB": 300.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 60.6, "isParetoFrontier": false }, { "id": "kimik25", "name": "Kimi K2.5", "org": "Moonshot", "releaseDate": "2026-01-27", "openWeights": true, "capability": 87.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.12, "FrontierMath-2025-02-28-Private": 0.279, "FrontierMath-Tier-4-2025-07-01-Private": 0.042, "GPQA diamond": 0.876, "OTIS Mock AIME 2024-2025": 0.922, "SWE-Bench verified": 0.7376, "SimpleQA Verified": 0.343 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 2.263947, "energyChatRange": [ 0.493201, 4.034693 ], "energyComplex": 52.825424, "energyComplexRange": [ 11.508013, 94.142836 ], "activeParamsB": [ 31.2, 1040.0 ], "totalParamsB": 1040.0, "sizeSource": "epoch_arch_unknown", "architectureKnown": false, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "claudesonnet46", "name": "Claude Sonnet 4.6", "org": "Anthropic", "releaseDate": "2026-02-17", "openWeights": false, "capability": 87.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.13, "FrontierMath-2025-02-28-Private": 0.324, "FrontierMath-2025-02-28-Public": 0.8, "FrontierMath-Tier-4-2025-07-01-Private": 0.083, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "GPQA diamond": 0.8737, "Mystery Game Puzzles": 0.16, "OTIS Mock AIME 2024-2025": 0.8579, "SWE-Bench verified": 0.7521, "SimpleQA Verified": 0.355 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": 0.32788, "energyChatRange": [ 0.250647, 0.405113 ], "energyComplex": 7.650538, "energyComplexRange": [ 5.84843, 9.452646 ], "activeParamsB": [ 44, 132 ], "totalParamsB": 440.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 41.4, "isParetoFrontier": false }, { "id": "qwen36max", "name": "Qwen 3.6 Max (Preview)", "org": "Alibaba", "releaseDate": "2026-04-20", "openWeights": false, "capability": 87.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.2, "FrontierMath-2025-02-28-Private": 0.231, "FrontierMath-2025-02-28-Public": 0.5, "FrontierMath-Tier-4-2025-07-01-Private": 0.0417, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "GPQA diamond": 0.8737, "Mystery Game Puzzles": 0.19, "OTIS Mock AIME 2024-2025": 0.9111, "SWE-Bench verified": 0.7665, "SimpleQA Verified": 0.52 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "grok4", "name": "Grok 4", "org": "xAI", "releaseDate": "2025-07-09", "openWeights": false, "capability": 87.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.28, "FrontierMath-2025-02-28-Private": 0.1966, "FrontierMath-Tier-4-2025-07-01-Private": 0.0208, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "GPQA diamond": 0.87, "OTIS Mock AIME 2024-2025": 0.84 }, "reasoning": { "status": "undisclosed", "note": "xAI publishes no position on whether reasoning text is returned, and no parameter counts either. Nothing here is assumed.", "source": null }, "energyChat": 23.230182, "energyChatRange": [ 2.798495, 43.661869 ], "energyComplex": 542.037586, "energyComplexRange": [ 65.298219, 1018.776953 ], "activeParamsB": [ 90.0, 3000.0 ], "totalParamsB": 3000.0, "sizeSource": "epoch_arch_unknown", "architectureKnown": false, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "qwen35397ba17b", "name": "Qwen3.5 397B-A17B", "org": "Alibaba", "releaseDate": "2026-02-13", "openWeights": true, "capability": 86.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.13, "FrontierMath-Tiers-1-3-v2-Private": 0.2947, "GPQA diamond": 0.8636, "Mystery Game Puzzles": 0.18, "OTIS Mock AIME 2024-2025": 0.8222 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.221675, "energyChatRange": [ 0.221675, 0.221675 ], "energyComplex": 5.172417, "energyComplexRange": [ 5.172417, 5.172417 ], "activeParamsB": 17.0, "totalParamsB": 397.0, "sizeSource": "epoch_active_in_name", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gpt5", "name": "GPT-5", "org": "OpenAI", "releaseDate": "2025-08-07", "openWeights": false, "capability": 85.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.37, "EBR-bench": 0.127, "FrontierMath-2025-02-28-Private": 0.2724, "FrontierMath-2025-02-28-Public": 0.7, "FrontierMath-Tier-4-2025-07-01-Private": 0.0625, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tier-4-v2-Private": 0.2195, "FrontierMath-Tiers-1-3-v2-Private": 0.5544, "GPQA diamond": 0.8535, "MATH level 5": 0.9792, "Mystery Game Puzzles": 0.23, "OTIS Mock AIME 2024-2025": 0.8722, "SWE-Bench verified": 0.7149, "SimpleQA Verified": 0.501 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.267698, "energyChatRange": [ 0.215039, 0.320357 ], "energyComplex": 6.246286, "energyComplexRange": [ 5.017576, 7.474996 ], "activeParamsB": [ 30, 90 ], "totalParamsB": 300.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 53.6, "isParetoFrontier": false }, { "id": "nemotron3ultra", "name": "Nemotron 3 Ultra", "org": "NVIDIA", "releaseDate": "2026-06-04", "openWeights": true, "capability": 85.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.12, "GPQA diamond": 0.8535, "Mystery Game Puzzles": 0.2, "OTIS Mock AIME 2024-2025": 0.8667 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.576753, "energyChatRange": [ 0.576753, 0.576753 ], "energyComplex": 13.457565, "energyComplexRange": [ 13.457565, 13.457565 ], "activeParamsB": 55.0, "totalParamsB": 550.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "qwen35plus", "name": "Qwen 3.5 Plus (hosted 397B-A17B)", "org": "Alibaba", "releaseDate": "2026-02-16", "openWeights": false, "capability": 84.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.22, "FrontierMath-2025-02-28-Private": 0.2103, "FrontierMath-2025-02-28-Public": 0.5, "FrontierMath-Tier-4-2025-07-01-Private": 0.0208, "GPQA diamond": 0.8485, "Mystery Game Puzzles": 0.16, "OTIS Mock AIME 2024-2025": 0.8667, "SimpleQA Verified": 0.2535 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "qwen3635ba3b", "name": "Qwen 3.6 35B-A3B", "org": "Alibaba", "releaseDate": "2026-04-14", "openWeights": true, "capability": 84.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.04, "FrontierMath-Tiers-1-3-v2-Private": 0.1754, "GPQA diamond": 0.8485, "Mystery Game Puzzles": 0.22, "OTIS Mock AIME 2024-2025": 0.6889 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.024638, "energyChatRange": [ 0.024638, 0.024638 ], "energyComplex": 0.574877, "energyComplexRange": [ 0.574877, 0.574877 ], "activeParamsB": 3.0, "totalParamsB": 35.0, "sizeSource": "epoch_active_in_name", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "qwen3627b", "name": "Qwen3.6 27B", "org": "Alibaba", "releaseDate": "2026-04-22", "openWeights": true, "capability": 84.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.09, "FrontierMath-Tiers-1-3-v2-Private": 0.3509, "GPQA diamond": 0.8485, "Mystery Game Puzzles": 0.07, "OTIS Mock AIME 2024-2025": 0.6667 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.015061, "energyChatRange": [ 0.015061, 0.015061 ], "energyComplex": 0.351434, "energyComplexRange": [ 0.351434, 0.351434 ], "activeParamsB": 28.0, "totalParamsB": 28.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": true }, { "id": "kimik2thinking", "name": "Kimi K2 Thinking", "org": "Moonshot", "releaseDate": "2025-11-06", "openWeights": true, "capability": 84.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.2, "FrontierMath-2025-02-28-Private": 0.214, "FrontierMath-2025-02-28-Public": 0.2, "FrontierMath-Tier-4-2025-07-01-Private": 0.0, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "GPQA diamond": 0.8422, "OTIS Mock AIME 2024-2025": 0.8306 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.496009, "energyChatRange": [ 0.496009, 0.496009 ], "energyComplex": 11.573544, "energyComplexRange": [ 11.573544, 11.573544 ], "activeParamsB": 32.0, "totalParamsB": 1000.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gemini25pro", "name": "Gemini 2.5 Pro (Mar 2025)", "org": "Google DeepMind", "releaseDate": "2025-03-31", "openWeights": false, "capability": 83.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.8384, "MATH level 5": 0.9556 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 3.397824, "energyChatRange": [ 1.993584, 4.802063 ], "energyComplex": 79.28255, "energyComplexRange": [ 46.516957, 112.048142 ], "activeParamsB": [ 200, 600 ], "totalParamsB": 2000.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 89.6, "isParetoFrontier": false }, { "id": "gpt54mini", "name": "GPT-5.4 Mini", "org": "OpenAI", "releaseDate": "2026-03-17", "openWeights": false, "capability": 83.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.18, "FrontierMath-2025-02-28-Private": 0.2828, "FrontierMath-2025-02-28-Public": 0.5, "FrontierMath-Tier-4-2025-07-01-Private": 0.0208, "FrontierMath-Tier-4-v2-Private": 0.0976, "FrontierMath-Tiers-1-3-v2-Private": 0.5123, "GPQA diamond": 0.8359, "Mystery Game Puzzles": 0.11, "OTIS Mock AIME 2024-2025": 0.8722, "SimpleQA Verified": 0.294 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.170123, "energyChatRange": [ 0.125363, 0.214883 ], "energyComplex": 3.969543, "energyComplexRange": [ 2.92514, 5.013946 ], "activeParamsB": [ 56, 158 ], "totalParamsB": 158.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 73.0, "isParetoFrontier": false }, { "id": "qwen3535ba3b", "name": "Qwen3.5-35B-A3B", "org": "Alibaba", "releaseDate": "2026-02-24", "openWeights": true, "capability": 83.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.01, "GPQA diamond": 0.8346, "OTIS Mock AIME 2024-2025": 0.5444 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.024638, "energyChatRange": [ 0.024638, 0.024638 ], "energyComplex": 0.574877, "energyComplexRange": [ 0.574877, 0.574877 ], "activeParamsB": 3.0, "totalParamsB": 35.0, "sizeSource": "epoch_active_in_name", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "deepseekv32", "name": "DeepSeek-V3.2", "org": "DeepSeek", "releaseDate": "2025-12-01", "openWeights": true, "capability": 83.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.14, "FrontierMath-2025-02-28-Private": 0.221, "FrontierMath-Tier-4-2025-07-01-Private": 0.021, "GPQA diamond": 0.8342, "OTIS Mock AIME 2024-2025": 0.8782 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "claudefable5", "name": "Claude Fable 5", "org": "Anthropic", "releaseDate": "2026-06-09", "openWeights": false, "capability": 83.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.41, "EBR-bench": 0.3952, "FrontierMath-2025-02-28-Public": 1.0, "FrontierMath-Tier-4-2025-07-01-Public": 1.0, "FrontierMath-Tier-4-v2-Private": 0.878, "FrontierMath-Tiers-1-3-v2-Private": 0.8702, "GPQA diamond": 0.8333, "MirrorCode": 0.6389, "Mystery Game Puzzles": 0.52, "OEIS Open Lite": 0.44, "OTIS Mock AIME 2024-2025": 0.9972, "SimpleQA Verified": 0.707 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "glm47", "name": "GLM-4.7", "org": "Z.ai (Zhipu AI)", "releaseDate": "2025-12-22", "openWeights": true, "capability": 83.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.06, "FrontierMath-2025-02-28-Private": 0.0244, "FrontierMath-Tier-4-2025-07-01-Private": 0.0, "GPQA diamond": 0.8333, "OTIS Mock AIME 2024-2025": 0.8333, "SimpleQA Verified": 0.322 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.820232, "energyChatRange": [ 0.820232, 0.820232 ], "energyComplex": 19.138751, "energyComplexRange": [ 19.138751, 19.138751 ], "activeParamsB": 358.0, "totalParamsB": 358.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gemini35flashlite", "name": "Gemini 3.5 Flash-Lite", "org": "Google DeepMind", "releaseDate": "2026-07-21", "openWeights": false, "capability": 83.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.22, "FrontierMath-Tier-4-v2-Private": 0.0, "FrontierMath-Tiers-1-3-v2-Private": 0.2596, "GPQA diamond": 0.8333, "Mystery Game Puzzles": 0.19, "OTIS Mock AIME 2024-2025": 0.7111 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "qwen36flash", "name": "Qwen 3.6 Flash", "org": "Alibaba", "releaseDate": "2026-04-27", "openWeights": false, "capability": 83.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.2, "FrontierMath-2025-02-28-Private": 0.1034, "FrontierMath-2025-02-28-Public": 0.2, "FrontierMath-Tier-4-2025-07-01-Private": 0.0, "FrontierMath-Tiers-1-3-v2-Private": 0.2246, "GPQA diamond": 0.8333, "Mystery Game Puzzles": 0.18, "OTIS Mock AIME 2024-2025": 0.8444, "SimpleQA Verified": 0.1592 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "gemini3flash", "name": "Gemini 3 Flash", "org": "Google DeepMind", "releaseDate": "2025-12-17", "openWeights": false, "capability": 83.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.38, "FrontierMath-2025-02-28-Private": 0.3564, "FrontierMath-2025-02-28-Public": 0.6, "FrontierMath-Tier-4-2025-07-01-Private": 0.0417, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tier-4-v2-Private": 0.1707, "FrontierMath-Tiers-1-3-v2-Private": 0.5123, "GPQA diamond": 0.8321, "Mystery Game Puzzles": 0.26, "OTIS Mock AIME 2024-2025": 0.9278, "SWE-Bench verified": 0.7541, "SimpleQA Verified": 0.668 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 0.503155, "energyChatRange": [ 0.327625, 0.678685 ], "energyComplex": 11.740273, "energyComplexRange": [ 7.644574, 15.835972 ], "activeParamsB": [ 100, 300 ], "totalParamsB": 300.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 73.9, "isParetoFrontier": false }, { "id": "gpt52", "name": "GPT-5.2", "org": "OpenAI", "releaseDate": "2025-12-11", "openWeights": false, "capability": 82.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.23, "EBR-bench": 0.2302, "FrontierMath-2025-02-28-Private": 0.2655, "FrontierMath-2025-02-28-Public": 0.4, "FrontierMath-Tier-4-2025-07-01-Private": 0.0625, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tier-4-v2-Private": 0.317, "FrontierMath-Tiers-1-3-v2-Private": 0.674, "GPQA diamond": 0.827, "Mystery Game Puzzles": 0.23, "OTIS Mock AIME 2024-2025": 0.7889, "SWE-Bench verified": 0.7376, "SimpleQA Verified": 0.371 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.318203, "energyChatRange": [ 0.24448, 0.391925 ], "energyComplex": 7.424731, "energyComplexRange": [ 5.704537, 9.144925 ], "activeParamsB": [ 42, 126 ], "totalParamsB": 420.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 43.8, "isParetoFrontier": false }, { "id": "gpt55instant", "name": "GPT-5.5 Instant", "org": "OpenAI", "releaseDate": "2026-05-05", "openWeights": false, "capability": 82.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.12, "FrontierMath-Tier-4-v2-Private": 0.0244, "FrontierMath-Tiers-1-3-v2-Private": 0.2632, "GPQA diamond": 0.8251, "OTIS Mock AIME 2024-2025": 0.6806 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "qwen35flash", "name": "Qwen 3.5 Flash (hosted 35B-A3B)", "org": "Alibaba", "releaseDate": "2026-02-25", "openWeights": false, "capability": 82.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.21, "FrontierMath-2025-02-28-Private": 0.0621, "FrontierMath-2025-02-28-Public": 0.1, "FrontierMath-Tier-4-2025-07-01-Private": 0.0, "FrontierMath-Tiers-1-3-v2-Private": 0.0947, "GPQA diamond": 0.8232, "Mystery Game Puzzles": 0.16, "OTIS Mock AIME 2024-2025": 0.8444, "SimpleQA Verified": 0.2032 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "qwen37plus", "name": "Qwen3.7-Plus", "org": "Alibaba", "releaseDate": "2026-06-02", "openWeights": false, "capability": 81.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.09, "FrontierMath-Tiers-1-3-v2-Private": 0.3439, "GPQA diamond": 0.8182, "Mystery Game Puzzles": 0.17, "OTIS Mock AIME 2024-2025": 0.8 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "o3", "name": "o3", "org": "OpenAI", "releaseDate": "2025-04-16", "openWeights": false, "capability": 81.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.27, "FrontierMath-2025-02-28-Private": 0.169, "FrontierMath-2025-02-28-Public": 0.1, "FrontierMath-Tier-4-2025-07-01-Private": 0.0208, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tiers-1-3-v2-Private": 0.193, "GPQA diamond": 0.8182, "MATH level 5": 0.9777, "Mystery Game Puzzles": 0.23, "OTIS Mock AIME 2024-2025": 0.8389, "SWE-Bench verified": 0.6232, "SimpleQA Verified": 0.494 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "qwen37flash", "name": "Qwen3.7 Flash", "org": "Alibaba", "releaseDate": "2026-07-27", "openWeights": false, "capability": 80.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.09, "FrontierMath-Tiers-1-3-v2-Private": 0.193, "GPQA diamond": 0.8081, "Mystery Game Puzzles": 0.14, "OTIS Mock AIME 2024-2025": 0.7778 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "claudeopus45", "name": "Claude Opus 4.5", "org": "Anthropic", "releaseDate": "2025-11-24", "openWeights": false, "capability": 80.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.12, "EBR-bench": 0.1429, "FrontierMath-2025-02-28-Private": 0.2034, "FrontierMath-2025-02-28-Public": 0.2, "FrontierMath-Tier-4-2025-07-01-Private": 0.0417, "FrontierMath-Tier-4-v2-Private": 0.0488, "FrontierMath-Tiers-1-3-v2-Private": 0.3439, "GPQA diamond": 0.8068, "Mystery Game Puzzles": 0.22, "OTIS Mock AIME 2024-2025": 0.8167, "SWE-Bench verified": 0.7665, "SimpleQA Verified": 0.457 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": 0.8199, "energyChatRange": [ 0.586445, 1.053354 ], "energyComplex": 19.130991, "energyComplexRange": [ 13.683712, 24.578271 ], "activeParamsB": [ 67, 200 ], "totalParamsB": 670.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 39.6, "isParetoFrontier": false }, { "id": "o4mini", "name": "o4-mini", "org": "OpenAI", "releaseDate": "2025-04-16", "openWeights": false, "capability": 79.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.26, "FrontierMath-2025-02-28-Private": 0.2483, "FrontierMath-2025-02-28-Public": 0.3, "FrontierMath-Tier-4-2025-07-01-Private": 0.0625, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tier-4-v2-Private": 0.0488, "FrontierMath-Tiers-1-3-v2-Private": 0.3614, "GPQA diamond": 0.7961, "MATH level 5": 0.9783, "Mystery Game Puzzles": 0.05, "OTIS Mock AIME 2024-2025": 0.8167, "SimpleQA Verified": 0.188 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.011095, "energyChatRange": [ 0.009998, 0.012192 ], "energyComplex": 0.258892, "energyComplexRange": [ 0.233293, 0.28449 ], "activeParamsB": [ 8, 28 ], "totalParamsB": 28.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 95.6, "isParetoFrontier": true }, { "id": "qwen359b", "name": "Qwen3.5-9B", "org": "Alibaba", "releaseDate": "2026-02-24", "openWeights": true, "capability": 79.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.02, "GPQA diamond": 0.7898, "OTIS Mock AIME 2024-2025": 0.4278 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.012977, "energyChatRange": [ 0.012977, 0.012977 ], "energyComplex": 0.302798, "energyComplexRange": [ 0.302798, 0.302798 ], "activeParamsB": 9.0, "totalParamsB": 9.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gpt54nano", "name": "GPT-5.4 Nano", "org": "OpenAI", "releaseDate": "2026-03-17", "openWeights": false, "capability": 78.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.3, "FrontierMath-2025-02-28-Private": 0.2586, "FrontierMath-2025-02-28-Public": 0.6, "FrontierMath-Tier-4-2025-07-01-Private": 0.0625, "FrontierMath-Tier-4-v2-Private": 0.122, "FrontierMath-Tiers-1-3-v2-Private": 0.4491, "GPQA diamond": 0.7847, "Mystery Game Puzzles": 0.09, "OTIS Mock AIME 2024-2025": 0.8778, "SimpleQA Verified": 0.117 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.027169, "energyChatRange": [ 0.022452, 0.031886 ], "energyComplex": 0.633944, "energyComplexRange": [ 0.523872, 0.744016 ], "activeParamsB": [ 15, 58 ], "totalParamsB": 58.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 70.8, "isParetoFrontier": false }, { "id": "grok3mini", "name": "Grok-3 mini", "org": "xAI", "releaseDate": "2025-04-09", "openWeights": false, "capability": 76.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0586, "FrontierMath-2025-02-28-Public": 0.0, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "GPQA diamond": 0.7626, "MATH level 5": 0.8807, "OTIS Mock AIME 2024-2025": 0.7778 }, "reasoning": { "status": "undisclosed", "note": "xAI publishes no position on whether reasoning text is returned, and no parameter counts either. Nothing here is assumed.", "source": null }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "gemma431b", "name": "Gemma 4 31B IT", "org": "Google DeepMind", "releaseDate": "2026-04-02", "openWeights": true, "capability": 75.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.05, "GPQA diamond": 0.7576, "OTIS Mock AIME 2024-2025": 0.7333, "SimpleQA Verified": 0.104 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 0.015391, "energyChatRange": [ 0.015391, 0.015391 ], "energyComplex": 0.359113, "energyComplexRange": [ 0.359113, 0.359113 ], "activeParamsB": 31.0, "totalParamsB": 31.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "grok3", "name": "Grok 3", "org": "xAI", "releaseDate": "2025-04-09", "openWeights": false, "capability": 75.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0379, "FrontierMath-2025-02-28-Public": 0.0, "FrontierMath-Tier-4-2025-07-01-Private": 0.0, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "GPQA diamond": 0.7576, "MATH level 5": 0.8875, "OTIS Mock AIME 2024-2025": 0.5556 }, "reasoning": { "status": "undisclosed", "note": "xAI publishes no position on whether reasoning text is returned, and no parameter counts either. Nothing here is assumed.", "source": null }, "energyChat": 23.230182, "energyChatRange": [ 2.798495, 43.661869 ], "energyComplex": 542.037586, "energyComplexRange": [ 65.298219, 1018.776953 ], "activeParamsB": [ 90.0, 3000.0 ], "totalParamsB": 3000.0, "sizeSource": "epoch_arch_unknown", "architectureKnown": false, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "gptoss120b", "name": "gpt-oss-120b", "org": "OpenAI", "releaseDate": "2025-08-05", "openWeights": true, "capability": 75.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.2, "GPQA diamond": 0.7576, "Mystery Game Puzzles": 0.02, "OTIS Mock AIME 2024-2025": 0.8889 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.050197, "energyChatRange": [ 0.050197, 0.050197 ], "energyComplex": 1.171257, "energyComplexRange": [ 1.171257, 1.171257 ], "activeParamsB": 5.1, "totalParamsB": 117.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "measured": { "energy": 0.127364, "tokens": 1776, "config": "B200 x1, batch 32", "task": "gpqa", "ratioToModel": 0.429, "source": "ML.ENERGY" }, "measurementDisagreement": true, "isParetoFrontier": false }, { "id": "o3mini", "name": "o3-mini", "org": "OpenAI", "releaseDate": "2025-01-31", "openWeights": false, "capability": 74.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.17, "FrontierMath-2025-02-28-Private": 0.0808, "FrontierMath-2025-02-28-Public": 0.1125, "FrontierMath-Tier-4-2025-07-01-Private": 0.0417, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tier-4-v2-Private": 0.0, "FrontierMath-Tiers-1-3-v2-Private": 0.186, "GPQA diamond": 0.7427, "MATH level 5": 0.9517, "Mystery Game Puzzles": 0.07, "OTIS Mock AIME 2024-2025": 0.6389, "SimpleQA Verified": 0.153 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.010519, "energyChatRange": [ 0.009421, 0.011616 ], "energyComplex": 0.245433, "energyComplexRange": [ 0.219834, 0.271031 ], "activeParamsB": [ 8, 28 ], "totalParamsB": 28.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 170.6, "isParetoFrontier": true }, { "id": "gemini31flashlite", "name": "Gemini 3.1 Flash-Lite", "org": "Google", "releaseDate": "2026-03-03", "openWeights": false, "capability": 74.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.2, "FrontierMath-Tiers-1-3-v2-Private": 0.214, "GPQA diamond": 0.7424, "OTIS Mock AIME 2024-2025": 0.8 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 0.066751, "energyChatRange": [ 0.050295, 0.083207 ], "energyComplex": 1.557519, "energyComplexRange": [ 1.173547, 1.94149 ], "activeParamsB": [ 30, 105 ], "totalParamsB": 105.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 85.1, "isParetoFrontier": false }, { "id": "claudesonnet45", "name": "Claude Sonnet 4.5", "org": "Anthropic", "releaseDate": "2025-09-29", "openWeights": false, "capability": 73.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.12, "EBR-bench": 0.0238, "FrontierMath-2025-02-28-Private": 0.1349, "FrontierMath-2025-02-28-Public": 0.0, "FrontierMath-Tier-4-2025-07-01-Private": 0.0208, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tier-4-v2-Private": 0.0244, "FrontierMath-Tiers-1-3-v2-Private": 0.2386, "GPQA diamond": 0.7374, "MATH level 5": 0.9773, "Mystery Game Puzzles": 0.17, "OTIS Mock AIME 2024-2025": 0.3556, "SWE-Bench verified": 0.7128, "SimpleQA Verified": 0.237 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": 0.341731, "energyChatRange": [ 0.264497, 0.418964 ], "energyComplex": 7.973715, "energyComplexRange": [ 6.171607, 9.775823 ], "activeParamsB": [ 44, 132 ], "totalParamsB": 440.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 32.2, "isParetoFrontier": false }, { "id": "claudeopus41", "name": "Claude Opus 4.1", "org": "Anthropic", "releaseDate": "2025-08-05", "openWeights": false, "capability": 73.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.07, "EBR-bench": 0.0794, "FrontierMath-2025-02-28-Private": 0.0586, "FrontierMath-2025-02-28-Public": 0.0, "FrontierMath-Tier-4-2025-07-01-Private": 0.0417, "FrontierMath-Tier-4-v2-Private": 0.0244, "FrontierMath-Tiers-1-3-v2-Private": 0.1263, "GPQA diamond": 0.7323, "Mystery Game Puzzles": 0.21, "OTIS Mock AIME 2024-2025": 0.4, "SWE-Bench verified": 0.7335 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": 3.971676, "energyChatRange": [ 2.567436, 5.375916 ], "energyComplex": 92.672442, "energyComplexRange": [ 59.90685, 125.438034 ], "activeParamsB": [ 200, 600 ], "totalParamsB": 2000.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 12.1, "isParetoFrontier": false }, { "id": "gemma426ba4b", "name": "Gemma 4 26B A4B", "org": "Google DeepMind", "releaseDate": "2026-04-02", "openWeights": true, "capability": 73.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.06, "GPQA diamond": 0.7323, "OTIS Mock AIME 2024-2025": 0.8222 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 0.010287, "energyChatRange": [ 0.010287, 0.010287 ], "energyComplex": 0.240029, "energyComplexRange": [ 0.240029, 0.240029 ], "activeParamsB": 4.0, "totalParamsB": 26.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 61.5, "isParetoFrontier": true }, { "id": "qwen3max", "name": "Qwen3-Max", "org": "Alibaba", "releaseDate": "2025-09-24", "openWeights": false, "capability": 72.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.04, "FrontierMath-Tiers-1-3-v2-Private": 0.1895, "GPQA diamond": 0.726, "MATH level 5": 0.9713, "Mystery Game Puzzles": 0.05, "OTIS Mock AIME 2024-2025": 0.7333, "SimpleQA Verified": 0.4875 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": 0.73473, "energyChatRange": [ 0.524094, 0.945366 ], "energyComplex": 17.143695, "energyComplexRange": [ 12.228856, 22.058533 ], "activeParamsB": [ 40.0, 160.0 ], "totalParamsB": 1000.0, "sizeSource": "epoch_moe_estimated", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "deepseekr1", "name": "DeepSeek-R1", "org": "DeepSeek", "releaseDate": "2025-01-20", "openWeights": true, "capability": 71.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.7172, "MATH level 5": 0.9305, "OTIS Mock AIME 2024-2025": 0.5333 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 2.739282, "energyChatRange": [ 2.739282, 2.739282 ], "energyComplex": 63.916577, "energyComplexRange": [ 63.916577, 63.916577 ], "activeParamsB": 671.0, "totalParamsB": 671.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gpt5mini", "name": "GPT-5 mini", "org": "OpenAI", "releaseDate": "2025-08-07", "openWeights": false, "capability": 71.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.12, "FrontierMath-2025-02-28-Private": 0.2034, "FrontierMath-2025-02-28-Public": 0.4, "FrontierMath-Tier-4-2025-07-01-Private": 0.0417, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tier-4-v2-Private": 0.122, "FrontierMath-Tiers-1-3-v2-Private": 0.4667, "GPQA diamond": 0.7165, "MATH level 5": 0.9679, "Mystery Game Puzzles": 0.1, "OTIS Mock AIME 2024-2025": 0.7833, "SWE-Bench verified": 0.6467, "SimpleQA Verified": 0.216 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.059145, "energyChatRange": [ 0.049272, 0.069019 ], "energyComplex": 1.380054, "energyComplexRange": [ 1.149671, 1.610437 ], "activeParamsB": [ 25, 70 ], "totalParamsB": 70.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 71.0, "isParetoFrontier": false }, { "id": "qwen3235ba22b", "name": "Qwen3-235B-A22B", "org": "Alibaba", "releaseDate": "2025-04-29", "openWeights": true, "capability": 70.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.7071, "MATH level 5": 0.6886 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.115226, "energyChatRange": [ 0.115226, 0.115226 ], "energyComplex": 2.688601, "energyComplexRange": [ 2.688601, 2.688601 ], "activeParamsB": 22.0, "totalParamsB": 235.0, "sizeSource": "epoch_active_in_name", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "claudeopus4", "name": "Claude Opus 4", "org": "Anthropic", "releaseDate": "2025-05-22", "openWeights": false, "capability": 69.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0448, "FrontierMath-2025-02-28-Public": 0.0, "FrontierMath-Tier-4-2025-07-01-Private": 0.0417, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "GPQA diamond": 0.6919, "MATH level 5": 0.8505, "OTIS Mock AIME 2024-2025": 0.4222, "SWE-Bench verified": 0.7066 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": 3.813116, "energyChatRange": [ 2.408877, 5.217356 ], "energyComplex": 88.972713, "energyComplexRange": [ 56.207121, 121.738306 ], "activeParamsB": [ 200, 600 ], "totalParamsB": 2000.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 15.9, "isParetoFrontier": false }, { "id": "gpt45", "name": "GPT-4.5", "org": "OpenAI", "releaseDate": "2025-02-27", "openWeights": false, "capability": 68.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.6869, "MATH level 5": 0.7863, "OTIS Mock AIME 2024-2025": 0.3778 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "gpt5nano", "name": "GPT-5 nano", "org": "OpenAI", "releaseDate": "2025-08-07", "openWeights": false, "capability": 67.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.01, "FrontierMath-2025-02-28-Private": 0.0724, "FrontierMath-2025-02-28-Public": 0.1, "FrontierMath-Tier-4-2025-07-01-Private": 0.0208, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tier-4-v2-Private": 0.0244, "FrontierMath-Tiers-1-3-v2-Private": 0.2, "GPQA diamond": 0.6742, "MATH level 5": 0.949, "Mystery Game Puzzles": 0.05, "OTIS Mock AIME 2024-2025": 0.7417, "SimpleQA Verified": 0.117 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.010403, "energyChatRange": [ 0.009662, 0.011143 ], "energyComplex": 0.242733, "energyComplexRange": [ 0.225455, 0.260012 ], "activeParamsB": [ 5, 18.5 ], "totalParamsB": 18.5, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 96.1, "isParetoFrontier": false }, { "id": "llama4maverick", "name": "Llama 4 Maverick", "org": "Meta AI", "releaseDate": "2025-04-05", "openWeights": true, "capability": 67.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0069, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.6698, "MATH level 5": 0.7302, "OTIS Mock AIME 2024-2025": 0.2056 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.221675, "energyChatRange": [ 0.221675, 0.221675 ], "energyComplex": 5.172417, "energyComplexRange": [ 5.172417, 5.172417 ], "activeParamsB": 17.0, "totalParamsB": 400.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gpt41", "name": "GPT-4.1", "org": "OpenAI", "releaseDate": "2025-04-14", "openWeights": false, "capability": 66.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.06, "FrontierMath-2025-02-28-Private": 0.0552, "FrontierMath-2025-02-28-Public": 0.0, "FrontierMath-Tier-4-2025-07-01-Private": 0.0, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "FrontierMath-Tiers-1-3-v2-Private": 0.0596, "GPQA diamond": 0.6692, "MATH level 5": 0.8301, "OTIS Mock AIME 2024-2025": 0.3833, "SWE-Bench verified": 0.4854, "SimpleQA Verified": 0.311 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.295908, "energyChatRange": [ 0.233595, 0.358221 ], "energyComplex": 6.904515, "energyComplexRange": [ 5.450542, 8.358488 ], "activeParamsB": [ 35, 106 ], "totalParamsB": 352.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 42.5, "isParetoFrontier": false }, { "id": "claudesonnet4", "name": "Claude Sonnet 4", "org": "Anthropic", "releaseDate": "2025-05-22", "openWeights": false, "capability": 66.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0414, "FrontierMath-2025-02-28-Public": 0.0, "FrontierMath-Tier-4-2025-07-01-Private": 0.0, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "GPQA diamond": 0.6667, "MATH level 5": 0.8437, "OTIS Mock AIME 2024-2025": 0.2889 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": 0.32908, "energyChatRange": [ 0.251847, 0.406313 ], "energyComplex": 7.678536, "energyComplexRange": [ 5.876428, 9.480644 ], "activeParamsB": [ 44, 132 ], "totalParamsB": 440.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 40.4, "isParetoFrontier": false }, { "id": "claude37sonnet", "name": "Claude 3.7 Sonnet", "org": "Anthropic", "releaseDate": "2025-02-24", "openWeights": false, "capability": 66.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.031, "FrontierMath-2025-02-28-Public": 0.0, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "GPQA diamond": 0.6604, "MATH level 5": 0.6818, "OTIS Mock AIME 2024-2025": 0.2194, "SWE-Bench verified": 0.6095 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "gpt41mini", "name": "GPT-4.1 mini", "org": "OpenAI", "releaseDate": "2025-04-14", "openWeights": false, "capability": 65.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.07, "FrontierMath-2025-02-28-Private": 0.0448, "FrontierMath-2025-02-28-Public": 0.1, "FrontierMath-Tiers-1-3-v2-Private": 0.0667, "GPQA diamond": 0.6585, "MATH level 5": 0.8729, "Mystery Game Puzzles": 0.07, "OTIS Mock AIME 2024-2025": 0.4472, "SimpleQA Verified": 0.127 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.07917, "energyChatRange": [ 0.063373, 0.094968 ], "energyComplex": 1.847306, "energyComplexRange": [ 1.478693, 2.215919 ], "activeParamsB": [ 40, 112 ], "totalParamsB": 112.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 34.4, "isParetoFrontier": false }, { "id": "gemini20pro", "name": "Gemini 2.0 Pro", "org": "Google DeepMind", "releaseDate": "2025-02-05", "openWeights": false, "capability": 65.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.6566, "MATH level 5": 0.8346 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "qwen332b", "name": "Qwen3-32B", "org": "Alibaba", "releaseDate": "2025-04-29", "openWeights": true, "capability": 65.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.05, "GPQA diamond": 0.6572, "OTIS Mock AIME 2024-2025": 0.6694 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.015588, "energyChatRange": [ 0.015588, 0.015588 ], "energyComplex": 0.363721, "energyComplexRange": [ 0.363721, 0.363721 ], "activeParamsB": 32.8, "totalParamsB": 32.8, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "measured": { "energy": 1.327115, "tokens": 6858, "config": "B200 x1, batch 32", "task": "gpqa", "ratioToModel": 3.725, "source": "ML.ENERGY" }, "measurementDisagreement": true, "isParetoFrontier": false }, { "id": "qwqplus", "name": "QWQ-Plus", "org": "Alibaba", "releaseDate": "2025-04-08", "openWeights": false, "capability": 65.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.654 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "qwq32b", "name": "QwQ-32B", "org": "Alibaba", "releaseDate": "2025-03-05", "openWeights": true, "capability": 65.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.05, "GPQA diamond": 0.6534, "OTIS Mock AIME 2024-2025": 0.5917 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.015555, "energyChatRange": [ 0.015555, 0.015555 ], "energyComplex": 0.362953, "energyComplexRange": [ 0.362953, 0.362953 ], "activeParamsB": 32.5, "totalParamsB": 32.5, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "deepseekr1distillqwen32b", "name": "DeepSeek-R1-Distill-Qwen-32B", "org": "DeepSeek", "releaseDate": "2025-01-20", "openWeights": true, "capability": 64.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.01, "GPQA diamond": 0.6414, "OTIS Mock AIME 2024-2025": 0.5556 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.0155, "energyChatRange": [ 0.0155, 0.0155 ], "energyComplex": 0.361673, "energyComplexRange": [ 0.361673, 0.361673 ], "activeParamsB": 32.0, "totalParamsB": 32.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gemini20flash", "name": "Gemini 2.0 Flash", "org": "Google DeepMind,Google", "releaseDate": "2025-02-05", "openWeights": false, "capability": 64.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0172, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.6414, "MATH level 5": 0.8217, "OTIS Mock AIME 2024-2025": 0.3111 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": 0.310471, "energyChatRange": [ 0.233237, 0.387704 ], "energyComplex": 7.244316, "energyComplexRange": [ 5.442208, 9.046423 ], "activeParamsB": [ 44, 132 ], "totalParamsB": 440.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 64.6, "isParetoFrontier": false }, { "id": "qwen314b", "name": "Qwen3-14B", "org": "Alibaba", "releaseDate": "2025-04-29", "openWeights": true, "capability": 63.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.04, "GPQA diamond": 0.6376, "OTIS Mock AIME 2024-2025": 0.6639 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.013613, "energyChatRange": [ 0.013613, 0.013613 ], "energyComplex": 0.317644, "energyComplexRange": [ 0.317644, 0.317644 ], "activeParamsB": 14.8, "totalParamsB": 14.8, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "measured": { "energy": 0.802118, "tokens": 6907, "config": "B200 x1, batch 32", "task": "gpqa", "ratioToModel": 2.559, "source": "ML.ENERGY" }, "measurementDisagreement": true, "isParetoFrontier": false }, { "id": "qwen330ba3b", "name": "Qwen3-30B-A3B", "org": "Alibaba", "releaseDate": "2025-04-29", "openWeights": true, "capability": 61.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.04, "GPQA diamond": 0.6168, "OTIS Mock AIME 2024-2025": 0.6278 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.012319, "energyChatRange": [ 0.012319, 0.012319 ], "energyComplex": 0.287439, "energyComplexRange": [ 0.287439, 0.287439 ], "activeParamsB": 3.0, "totalParamsB": 30.0, "sizeSource": "epoch_active_in_name", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "claudehaiku45", "name": "Claude Haiku 4.5", "org": "Anthropic", "releaseDate": "2025-10-15", "openWeights": false, "capability": 60.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.08, "FrontierMath-2025-02-28-Private": 0.0414, "FrontierMath-2025-02-28-Public": 0.0, "FrontierMath-Tier-4-2025-07-01-Private": 0.0208, "GPQA diamond": 0.6048, "MATH level 5": 0.8691, "OTIS Mock AIME 2024-2025": 0.3583, "SimpleQA Verified": 0.132 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": 0.024082, "energyChatRange": [ 0.02134, 0.026825 ], "energyComplex": 0.56192, "energyComplexRange": [ 0.497925, 0.625916 ], "activeParamsB": [ 10, 35 ], "totalParamsB": 35.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 71.1, "isParetoFrontier": false }, { "id": "mistralmedium3", "name": "Mistral Medium 3", "org": "Mistral AI", "releaseDate": "2025-05-07", "openWeights": false, "capability": 59.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0035, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.5953, "MATH level 5": 0.8163, "OTIS Mock AIME 2024-2025": 0.3222 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": 0.101947, "energyChatRange": [ 0.101947, 0.101947 ], "energyComplex": 2.37876, "energyComplexRange": [ 2.37876, 2.37876 ], "activeParamsB": 128.0, "totalParamsB": 128.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 34.5, "isParetoFrontier": false }, { "id": "o1mini", "name": "o1-mini", "org": "OpenAI", "releaseDate": "2024-09-12", "openWeights": false, "capability": 59.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0138, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.595, "MATH level 5": 0.8425, "OTIS Mock AIME 2024-2025": 0.4472 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "gemini15pro", "name": "Gemini 1.5 Pro", "org": "Google DeepMind", "releaseDate": "2024-05-14", "openWeights": false, "capability": 57.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.5723, "MATH level 5": 0.4075, "OTIS Mock AIME 2024-2025": 0.2306 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "gemini20flashthinking", "name": "Gemini 2.0 Flash Thinking", "org": "Google DeepMind,Google", "releaseDate": "2025-01-21", "openWeights": false, "capability": 57.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.5707, "OTIS Mock AIME 2024-2025": 0.5778 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "qwen38b", "name": "Qwen3-8B", "org": "Alibaba", "releaseDate": "2025-04-28", "openWeights": true, "capability": 56.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.5676, "OTIS Mock AIME 2024-2025": 0.5611 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.012889, "energyChatRange": [ 0.012889, 0.012889 ], "energyComplex": 0.30075, "energyComplexRange": [ 0.30075, 0.30075 ], "activeParamsB": 8.2, "totalParamsB": 8.2, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "measured": { "energy": 0.867311, "tokens": 7911, "config": "B200 x1, batch 32", "task": "gpqa", "ratioToModel": 2.552, "source": "ML.ENERGY" }, "measurementDisagreement": true, "isParetoFrontier": false }, { "id": "deepseekv3", "name": "DeepSeek-V3", "org": "DeepSeek", "releaseDate": "2024-12-26", "openWeights": true, "capability": 56.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0172, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.5653, "MATH level 5": 0.6485, "OTIS Mock AIME 2024-2025": 0.1583 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.513562, "energyChatRange": [ 0.513562, 0.513562 ], "energyComplex": 11.983114, "energyComplexRange": [ 11.983114, 11.983114 ], "activeParamsB": 37.0, "totalParamsB": 671.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "magistralsmall10", "name": "Magistral Small 1.0", "org": "Mistral AI", "releaseDate": "2025-06-10", "openWeights": true, "capability": 56.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.5606, "OTIS Mock AIME 2024-2025": 0.3 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.014623, "energyChatRange": [ 0.014623, 0.014623 ], "energyComplex": 0.341195, "energyComplexRange": [ 0.341195, 0.341195 ], "activeParamsB": 24.0, "totalParamsB": 24.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "phi4", "name": "Phi-4", "org": "Microsoft Research", "releaseDate": "2024-12-12", "openWeights": true, "capability": 56.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.01, "GPQA diamond": 0.5606, "MATH level 5": 0.6494, "OTIS Mock AIME 2024-2025": 0.1375 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.013526, "energyChatRange": [ 0.013526, 0.013526 ], "energyComplex": 0.315597, "energyComplexRange": [ 0.315597, 0.315597 ], "activeParamsB": 14.0, "totalParamsB": 14.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "qwen25max", "name": "Qwen2.5-Max", "org": "Alibaba", "releaseDate": "2025-01-25", "openWeights": false, "capability": 56.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0103, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.5612, "MATH level 5": 0.6718, "OTIS Mock AIME 2024-2025": 0.1611 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "deepseekr1distillllama70b", "name": "DeepSeek-R1-Distill-Llama-70B", "org": "DeepSeek", "releaseDate": "2025-01-20", "openWeights": true, "capability": 55.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.5574, "MATH level 5": 0.899, "OTIS Mock AIME 2024-2025": 0.5139 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.078676, "energyChatRange": [ 0.078676, 0.078676 ], "energyComplex": 1.835784, "energyComplexRange": [ 1.835784, 1.835784 ], "activeParamsB": 70.0, "totalParamsB": 70.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "claude35sonnet", "name": "Claude 3.5 Sonnet", "org": "Anthropic", "releaseDate": "2024-06-20", "openWeights": false, "capability": 54.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0103, "FrontierMath-2025-02-28-Public": 0.0, "FrontierMath-Tier-4-2025-07-01-Private": 0.0, "FrontierMath-Tier-4-2025-07-01-Public": 0.0, "GPQA diamond": 0.5404, "MATH level 5": 0.5168, "OTIS Mock AIME 2024-2025": 0.0653 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "grok2", "name": "Grok-2", "org": "xAI", "releaseDate": "2024-12-12", "openWeights": false, "capability": 53.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0069, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.5379, "MATH level 5": 0.6352, "OTIS Mock AIME 2024-2025": 0.1153 }, "reasoning": { "status": "undisclosed", "note": "xAI publishes no position on whether reasoning text is returned, and no parameter counts either. Nothing here is assumed.", "source": null }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "qwen34b", "name": "Qwen3-4B", "org": "Alibaba", "releaseDate": "2025-04-29", "openWeights": true, "capability": 52.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.01, "GPQA diamond": 0.5234 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.012429, "energyChatRange": [ 0.012429, 0.012429 ], "energyComplex": 0.289999, "energyComplexRange": [ 0.289999, 0.289999 ], "activeParamsB": 4.0, "totalParamsB": 4.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "llama4scout", "name": "Llama 4 Scout", "org": "Meta AI", "releaseDate": "2025-04-05", "openWeights": true, "capability": 51.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.5183, "MATH level 5": 0.6227, "OTIS Mock AIME 2024-2025": 0.0778 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.055419, "energyChatRange": [ 0.055419, 0.055419 ], "energyComplex": 1.293104, "energyComplexRange": [ 1.293104, 1.293104 ], "activeParamsB": 17.0, "totalParamsB": 109.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "llama31405b", "name": "Llama 3.1-405B", "org": "Meta AI", "releaseDate": "2024-07-23", "openWeights": true, "capability": 50.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.5092, "MATH level 5": 0.4977, "OTIS Mock AIME 2024-2025": 0.0972 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.557947, "energyChatRange": [ 0.213162, 0.902731 ], "energyComplex": 13.018752, "energyComplexRange": [ 4.973776, 21.063729 ], "activeParamsB": [ 12.15, 405.0 ], "totalParamsB": 405.0, "sizeSource": "epoch_arch_unknown", "architectureKnown": false, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "measured": { "energy": 0.646218, "tokens": 364, "config": "B200 x8, batch 32", "task": "lm-arena-chat", "ratioToModel": 0.956, "source": "ML.ENERGY" }, "measurementDisagreement": false, "isParetoFrontier": false }, { "id": "o1", "name": "o1-preview", "org": "OpenAI", "releaseDate": "2024-09-12", "openWeights": false, "capability": 50.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.5032, "MATH level 5": 0.8165, "OTIS Mock AIME 2024-2025": 0.3111 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.294096, "energyChatRange": [ 0.216862, 0.371329 ], "energyComplex": 6.862232, "energyComplexRange": [ 5.060125, 8.66434 ], "activeParamsB": [ 44, 132 ], "totalParamsB": 440.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 136.6, "isParetoFrontier": false }, { "id": "mistralsmall32", "name": "Mistral Small 3.2", "org": "Mistral AI", "releaseDate": "2025-06-20", "openWeights": true, "capability": 49.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.01, "GPQA diamond": 0.4905, "OTIS Mock AIME 2024-2025": 0.3028 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.014623, "energyChatRange": [ 0.014623, 0.014623 ], "energyComplex": 0.341195, "energyComplexRange": [ 0.341195, 0.341195 ], "activeParamsB": 24.0, "totalParamsB": 24.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "qwen2572b", "name": "Qwen2.5-72B", "org": "Alibaba", "releaseDate": "2024-09-19", "openWeights": true, "capability": 49.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.4915, "MATH level 5": 0.6317, "OTIS Mock AIME 2024-2025": 0.0806 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.079861, "energyChatRange": [ 0.079861, 0.079861 ], "energyComplex": 1.86343, "energyComplexRange": [ 1.86343, 1.86343 ], "activeParamsB": 72.7, "totalParamsB": 72.7, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "mistrallarge2", "name": "Mistral Large 2", "org": "Mistral AI", "releaseDate": "2024-11-18", "openWeights": true, "capability": 49.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0034, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.4902, "MATH level 5": 0.4482, "OTIS Mock AIME 2024-2025": 0.0847 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.075756, "energyChatRange": [ 0.049578, 0.101934 ], "energyComplex": 1.767642, "energyComplexRange": [ 1.15682, 2.378465 ], "activeParamsB": [ 3.69, 123.0 ], "totalParamsB": 123.0, "sizeSource": "epoch_arch_unknown", "architectureKnown": false, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "gpt41nano", "name": "GPT-4.1 nano", "org": "OpenAI", "releaseDate": "2025-04-14", "openWeights": false, "capability": 48.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0103, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.4893, "MATH level 5": 0.7, "OTIS Mock AIME 2024-2025": 0.2889, "SimpleQA Verified": 0.06 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.030385, "energyChatRange": [ 0.027423, 0.033347 ], "energyComplex": 0.708985, "energyComplexRange": [ 0.63987, 0.7781 ], "activeParamsB": [ 10, 37 ], "totalParamsB": 37.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 26.1, "isParetoFrontier": false }, { "id": "gpt4o", "name": "GPT-4o", "org": "OpenAI", "releaseDate": "2024-08-06", "openWeights": false, "capability": 48.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.13, "FrontierMath-2025-02-28-Private": 0.0034, "FrontierMath-2025-02-28-Public": 0.0, "FrontierMath-Tiers-1-3-v2-Private": 0.0035, "GPQA diamond": 0.489, "MATH level 5": 0.5328, "OTIS Mock AIME 2024-2025": 0.0625, "SWE-Bench verified": 0.3099, "SimpleQA Verified": 0.26 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.322844, "energyChatRange": [ 0.24561, 0.400077 ], "energyComplex": 7.533019, "energyComplexRange": [ 5.730912, 9.335127 ], "activeParamsB": [ 44, 132 ], "totalParamsB": 440.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 46.2, "isParetoFrontier": false }, { "id": "qwenplus", "name": "Qwen Plus", "org": "Alibaba", "releaseDate": "2025-04-28", "openWeights": false, "capability": 48.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0172, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.4811, "MATH level 5": 0.6528, "OTIS Mock AIME 2024-2025": 0.1778 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "gemma327b", "name": "Gemma 3 27B", "org": "Google DeepMind", "releaseDate": "2025-03-12", "openWeights": true, "capability": 47.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.4773, "MATH level 5": 0.7404, "OTIS Mock AIME 2024-2025": 0.225 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 0.014952, "energyChatRange": [ 0.014952, 0.014952 ], "energyComplex": 0.348874, "energyComplexRange": [ 0.348874, 0.348874 ], "activeParamsB": 27.0, "totalParamsB": 27.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "measured": { "energy": 0.092916, "tokens": 717, "config": "B200 x1, batch 32", "task": "lm-arena-chat", "ratioToModel": 2.599, "source": "ML.ENERGY" }, "measurementDisagreement": true, "isParetoFrontier": false }, { "id": "magistralsmall12", "name": "Magistral Small 1.2", "org": "Mistral AI", "releaseDate": "2025-09-18", "openWeights": true, "capability": 47.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.03, "GPQA diamond": 0.476, "OTIS Mock AIME 2024-2025": 0.2806 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.014623, "energyChatRange": [ 0.014623, 0.014623 ], "energyComplex": 0.341195, "energyComplexRange": [ 0.341195, 0.341195 ], "activeParamsB": 24.0, "totalParamsB": 24.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "mistralsmall31", "name": "Mistral Small 3.1", "org": "Mistral AI", "releaseDate": "2025-03-17", "openWeights": true, "capability": 47.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.01, "GPQA diamond": 0.4747, "MATH level 5": 0.4677, "OTIS Mock AIME 2024-2025": 0.0583 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.014623, "energyChatRange": [ 0.014623, 0.014623 ], "energyComplex": 0.341195, "energyComplexRange": [ 0.341195, 0.341195 ], "activeParamsB": 24.0, "totalParamsB": 24.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "llama3370b", "name": "Llama 3.3 70B", "org": "Meta AI", "releaseDate": "2024-12-06", "openWeights": true, "capability": 47.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.4744, "MATH level 5": 0.416, "OTIS Mock AIME 2024-2025": 0.0514 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.078676, "energyChatRange": [ 0.078676, 0.078676 ], "energyComplex": 1.835784, "energyComplexRange": [ 1.835784, 1.835784 ], "activeParamsB": 70.0, "totalParamsB": 70.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "measured": { "energy": 0.124099, "tokens": 407, "config": "B200 x2, batch 32", "task": "lm-arena-chat", "ratioToModel": 1.163, "source": "ML.ENERGY" }, "measurementDisagreement": false, "isParetoFrontier": false }, { "id": "gemini15flash", "name": "Gemini 1.5 Flash", "org": "Google DeepMind", "releaseDate": "2024-09-24", "openWeights": false, "capability": 47.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.4732, "MATH level 5": 0.6187, "OTIS Mock AIME 2024-2025": 0.0389 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "claude3opus", "name": "Claude 3 Opus", "org": "Anthropic", "releaseDate": "2024-02-29", "openWeights": false, "capability": 47.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.05, "GPQA diamond": 0.4716, "MATH level 5": 0.3748, "OTIS Mock AIME 2024-2025": 0.0472, "SimpleQA Verified": 0.126 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "tulu370b", "name": "Tulu 3 (T\u00fclu 3) 70B", "org": "Allen Institute for AI,University of Washington", "releaseDate": "2024-11-21", "openWeights": true, "capability": 46.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.4628, "MATH level 5": 0.4266, "OTIS Mock AIME 2024-2025": 0.0444 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.078676, "energyChatRange": [ 0.078676, 0.078676 ], "energyComplex": 1.835784, "energyComplexRange": [ 1.835784, 1.835784 ], "activeParamsB": 70.0, "totalParamsB": 70.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "qwen2532b", "name": "Qwen2.5-32B", "org": "Alibaba", "releaseDate": "2024-09-17", "openWeights": true, "capability": 46.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.4609, "MATH level 5": 0.5607, "OTIS Mock AIME 2024-2025": 0.0736 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.015555, "energyChatRange": [ 0.015555, 0.015555 ], "energyComplex": 0.362953, "energyComplexRange": [ 0.362953, 0.362953 ], "activeParamsB": 32.5, "totalParamsB": 32.5, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gptoss20b", "name": "gpt-oss-20b", "org": "OpenAI", "releaseDate": "2025-08-05", "openWeights": true, "capability": 46.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.02, "GPQA diamond": 0.4596, "OTIS Mock AIME 2024-2025": 0.4028 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.012385, "energyChatRange": [ 0.012385, 0.012385 ], "energyComplex": 0.288975, "energyComplexRange": [ 0.288975, 0.288975 ], "activeParamsB": 3.6, "totalParamsB": 21.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "measured": { "energy": 0.06461, "tokens": 1979, "config": "B200 x1, batch 32", "task": "gpqa", "ratioToModel": 0.791, "source": "ML.ENERGY" }, "measurementDisagreement": false, "isParetoFrontier": false }, { "id": "mistralsmall3", "name": "Mistral Small 3", "org": "Mistral AI", "releaseDate": "2025-01-30", "openWeights": true, "capability": 45.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.453, "MATH level 5": 0.4482, "OTIS Mock AIME 2024-2025": 0.0528 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.014623, "energyChatRange": [ 0.014623, 0.014623 ], "energyComplex": 0.341195, "energyComplexRange": [ 0.341195, 0.341195 ], "activeParamsB": 24.0, "totalParamsB": 24.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "deepseekr1distillqwen14b", "name": "DeepSeek-R1-Distill-Qwen-14B", "org": "DeepSeek", "releaseDate": "2025-01-20", "openWeights": true, "capability": 44.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.01, "GPQA diamond": 0.447, "MATH level 5": 0.8712, "OTIS Mock AIME 2024-2025": 0.5056 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.013613, "energyChatRange": [ 0.013613, 0.013613 ], "energyComplex": 0.317644, "energyComplexRange": [ 0.317644, 0.317644 ], "activeParamsB": 14.8, "totalParamsB": 14.8, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "llama3170b", "name": "Llama 3.1-70B", "org": "Meta AI", "releaseDate": "2024-07-23", "openWeights": true, "capability": 44.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.4419, "MATH level 5": 0.3668, "OTIS Mock AIME 2024-2025": 0.0361 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.078676, "energyChatRange": [ 0.078676, 0.078676 ], "energyComplex": 1.835784, "energyComplexRange": [ 1.835784, 1.835784 ], "activeParamsB": 70.0, "totalParamsB": 70.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "measured": { "energy": 0.104956, "tokens": 346, "config": "B200 x2, batch 32", "task": "lm-arena-chat", "ratioToModel": 1.156, "source": "ML.ENERGY" }, "measurementDisagreement": false, "isParetoFrontier": false }, { "id": "wizardlm28x22b", "name": "WizardLM-2 8x22B", "org": "Microsoft", "releaseDate": "2024-04-15", "openWeights": true, "capability": 43.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.4343, "MATH level 5": 0.2574 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.108292, "energyChatRange": [ 0.100867, 0.115717 ], "energyComplex": 2.526821, "energyComplexRange": [ 2.353573, 2.700069 ], "activeParamsB": [ 5.64, 22.56 ], "totalParamsB": 141.0, "sizeSource": "epoch_moe_estimated", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "gpt4turbo", "name": "GPT-4 Turbo (Nov 2023)", "org": "OpenAI", "releaseDate": "2024-01-25", "openWeights": false, "capability": 42.4, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.4236, "MATH level 5": 0.3541 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 1.137126, "energyChatRange": [ 0.828194, 1.446059 ], "energyComplex": 26.532951, "energyComplexRange": [ 19.324521, 33.741381 ], "activeParamsB": [ 88, 264 ], "totalParamsB": 880.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 14.9, "isParetoFrontier": false }, { "id": "qwenturbo", "name": "Qwen-Turbo", "org": "Alibaba", "releaseDate": "2024-11-01", "openWeights": false, "capability": 41.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.4179, "MATH level 5": 0.5623, "OTIS Mock AIME 2024-2025": 0.0611 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "llama3290b", "name": "Llama 3.2 90B", "org": "Meta AI", "releaseDate": "2024-09-24", "openWeights": true, "capability": 41.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.4104, "MATH level 5": 0.3944, "OTIS Mock AIME 2024-2025": 0.0264 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.086839, "energyChatRange": [ 0.086839, 0.086839 ], "energyComplex": 2.026234, "energyComplexRange": [ 2.026234, 2.026234 ], "activeParamsB": 88.6, "totalParamsB": 88.6, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "qwen272b", "name": "Qwen2-72B", "org": "Alibaba", "releaseDate": "2024-06-07", "openWeights": true, "capability": 40.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.4078, "MATH level 5": 0.3907 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.079866, "energyChatRange": [ 0.079866, 0.079866 ], "energyComplex": 1.863533, "energyComplexRange": [ 1.863533, 1.863533 ], "activeParamsB": 72.71, "totalParamsB": 72.71, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "claude3sonnet", "name": "Claude 3 Sonnet", "org": "Anthropic", "releaseDate": "2024-02-29", "openWeights": false, "capability": 40.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.4059, "MATH level 5": 0.1817, "OTIS Mock AIME 2024-2025": 0.025 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "llama370b", "name": "Llama 3-70B", "org": "Meta AI", "releaseDate": "2024-04-18", "openWeights": true, "capability": 40.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.4056, "MATH level 5": 0.2255, "OTIS Mock AIME 2024-2025": 0.0431 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.078676, "energyChatRange": [ 0.078676, 0.078676 ], "energyComplex": 1.835784, "energyComplexRange": [ 1.835784, 1.835784 ], "activeParamsB": 70.0, "totalParamsB": 70.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gemma312b", "name": "Gemma 3 12B", "org": "Google DeepMind", "releaseDate": "2025-03-12", "openWeights": true, "capability": 39.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.3946, "OTIS Mock AIME 2024-2025": 0.1667 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 0.013306, "energyChatRange": [ 0.013306, 0.013306 ], "energyComplex": 0.310477, "energyComplexRange": [ 0.310477, 0.310477 ], "activeParamsB": 12.0, "totalParamsB": 12.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "measured": { "energy": 0.049679, "tokens": 725, "config": "B200 x1, batch 32", "task": "lm-arena-chat", "ratioToModel": 1.544, "source": "ML.ENERGY" }, "measurementDisagreement": false, "isParetoFrontier": false }, { "id": "mistrallarge", "name": "Mistral Large", "org": "Mistral AI", "releaseDate": "2024-02-26", "openWeights": false, "capability": 38.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.3876, "MATH level 5": 0.2446, "OTIS Mock AIME 2024-2025": 0.0194 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": 0.099303, "energyChatRange": [ 0.099303, 0.099303 ], "energyComplex": 2.317079, "energyComplexRange": [ 2.317079, 2.317079 ], "activeParamsB": 123.0, "totalParamsB": 123.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "speed": 35.6, "isParetoFrontier": false }, { "id": "claude35haiku", "name": "Claude 3.5 Haiku", "org": "Anthropic", "releaseDate": "2024-10-22", "openWeights": false, "capability": 38.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "FrontierMath-2025-02-28-Private": 0.0034, "FrontierMath-2025-02-28-Public": 0.0, "GPQA diamond": 0.3813, "MATH level 5": 0.4636, "OTIS Mock AIME 2024-2025": 0.0431 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "qwen317b", "name": "Qwen3-1.7B", "org": "Alibaba", "releaseDate": "2025-04-29", "openWeights": true, "capability": 38.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.3801, "OTIS Mock AIME 2024-2025": 0.0806 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.012176, "energyChatRange": [ 0.012176, 0.012176 ], "energyComplex": 0.284111, "energyComplexRange": [ 0.284111, 0.284111 ], "activeParamsB": 1.7, "totalParamsB": 1.7, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gpt4omini", "name": "GPT-4o mini", "org": "OpenAI", "releaseDate": "2024-07-18", "openWeights": false, "capability": 37.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "FrontierMath-Tiers-1-3-v2-Private": 0.007, "GPQA diamond": 0.3772, "MATH level 5": 0.5263, "Mystery Game Puzzles": 0.12, "OTIS Mock AIME 2024-2025": 0.0694, "SimpleQA Verified": 0.083 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.013584, "energyChatRange": [ 0.012487, 0.014681 ], "energyComplex": 0.316967, "energyComplexRange": [ 0.291369, 0.342565 ], "activeParamsB": [ 8, 28 ], "totalParamsB": 28.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 33.0, "isParetoFrontier": false }, { "id": "hermes2thetallama370b", "name": "Hermes 2 Theta Llama-3 70B", "org": "Nous Research,Arcee AI", "releaseDate": "2024-06-20", "openWeights": true, "capability": 37.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.3747, "MATH level 5": 0.2269, "OTIS Mock AIME 2024-2025": 0.025 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.078676, "energyChatRange": [ 0.078676, 0.078676 ], "energyComplex": 1.835784, "energyComplexRange": [ 1.835784, 1.835784 ], "activeParamsB": 70.0, "totalParamsB": 70.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gemma227b", "name": "Gemma 2 27B", "org": "Google DeepMind", "releaseDate": "2024-06-24", "openWeights": true, "capability": 36.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.3649, "MATH level 5": 0.2789, "OTIS Mock AIME 2024-2025": 0.0139 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 0.014977, "energyChatRange": [ 0.014977, 0.014977 ], "energyComplex": 0.349463, "energyComplexRange": [ 0.349463, 0.349463 ], "activeParamsB": 27.23, "totalParamsB": 27.23, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "claude3haiku", "name": "Claude 3 Haiku", "org": "Anthropic", "releaseDate": "2024-03-07", "openWeights": false, "capability": 36.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.363, "MATH level 5": 0.1488, "OTIS Mock AIME 2024-2025": 0.0181 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "gpt4", "name": "GPT-4 (Mar 2023)", "org": "OpenAI", "releaseDate": "2023-03-14", "openWeights": false, "capability": 35.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.3573, "OTIS Mock AIME 2024-2025": 0.0056 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 3.329591, "energyChatRange": [ 2.09386, 4.565322 ], "energyComplex": 77.69046, "energyComplexRange": [ 48.856739, 106.524181 ], "activeParamsB": [ 176, 528 ], "totalParamsB": 1760.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 22.4, "isParetoFrontier": false }, { "id": "qwen257b", "name": "Qwen2.5-7B", "org": "Alibaba", "releaseDate": "2025-02-26", "openWeights": true, "capability": 35.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.3548, "OTIS Mock AIME 2024-2025": 0.025 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.012825, "energyChatRange": [ 0.012825, 0.012825 ], "energyComplex": 0.299239, "energyComplexRange": [ 0.299239, 0.299239 ], "activeParamsB": 7.61, "totalParamsB": 7.61, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "claude2", "name": "Claude 2", "org": "Anthropic", "releaseDate": "2023-07-11", "openWeights": false, "capability": 34.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.3466, "MATH level 5": 0.1173, "OTIS Mock AIME 2024-2025": 0.025 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "mixtral8x22b", "name": "Mixtral 8x22B", "org": "Mistral AI", "releaseDate": "2024-04-17", "openWeights": true, "capability": 34.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.3406, "MATH level 5": 0.2424 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.219666, "energyChatRange": [ 0.219666, 0.219666 ], "energyComplex": 5.125542, "energyComplexRange": [ 5.125542, 5.125542 ], "activeParamsB": 141.0, "totalParamsB": 141.0, "sizeSource": "epoch_active_in_notes", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gemini10pro", "name": "Gemini 1.0 Pro", "org": "Google DeepMind", "releaseDate": "2023-12-13", "openWeights": false, "capability": 34.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.3396, "MATH level 5": 0.1124, "OTIS Mock AIME 2024-2025": 0.0111 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "eurus27bprime", "name": "Eurus-2-7B-PRIME", "org": "Tsinghua University,University of Illinois Urbana-Champaign (UIUC),Shanghai AI Lab,Peking University,Shanghai Jiao Tong University,CUHK Shenzhen Research Institute", "releaseDate": "2024-12-31", "openWeights": true, "capability": 33.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.339 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.012758, "energyChatRange": [ 0.012758, 0.012758 ], "energyComplex": 0.297678, "energyComplexRange": [ 0.297678, 0.297678 ], "activeParamsB": 7.0, "totalParamsB": 7.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "deepseekr1distillqwen15b", "name": "DeepSeek-R1-Distill-Qwen-1.5B", "org": "DeepSeek", "releaseDate": "2025-01-20", "openWeights": true, "capability": 33.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.3359, "OTIS Mock AIME 2024-2025": 0.2139 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.012185, "energyChatRange": [ 0.012185, 0.012185 ], "energyComplex": 0.284316, "energyComplexRange": [ 0.284316, 0.284316 ], "activeParamsB": 1.78, "totalParamsB": 1.78, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "claude21", "name": "Claude 2.1", "org": "Anthropic", "releaseDate": "2023-11-21", "openWeights": false, "capability": 33.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.3295, "OTIS Mock AIME 2024-2025": 0.0194 }, "reasoning": { "status": "hidden", "note": "Thinking is returned in summarised form. Billing counts the full raw thinking generated, which is longer than the summary shown.", "source": "https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking" }, "energyChat": null, "energyComplex": null, "energySource": "unknown", "energyNote": "No parameter count published by any source yet." }, { "id": "gemini15flash8b", "name": "Gemini 1.5 Flash 8B", "org": "Google DeepMind", "releaseDate": "2024-10-03", "openWeights": false, "capability": 33.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.3295, "OTIS Mock AIME 2024-2025": 0.0458 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 0.012867, "energyChatRange": [ 0.012867, 0.012867 ], "energyComplex": 0.300238, "energyComplexRange": [ 0.300238, 0.300238 ], "activeParamsB": 8.0, "totalParamsB": 8.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "dbrx", "name": "DBRX", "org": "Databricks", "releaseDate": "2024-03-27", "openWeights": true, "capability": 32.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.3289, "MATH level 5": 0.1165 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.063756, "energyChatRange": [ 0.063756, 0.063756 ], "energyComplex": 1.48765, "energyComplexRange": [ 1.48765, 1.48765 ], "activeParamsB": 36.0, "totalParamsB": 132.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "yi1534b", "name": "Yi-1.5-34B", "org": "01.AI", "releaseDate": "2024-05-13", "openWeights": true, "capability": 32.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.3198, "MATH level 5": 0.2548 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.031439, "energyChatRange": [ 0.031439, 0.031439 ], "energyComplex": 0.733586, "energyComplexRange": [ 0.733586, 0.733586 ], "activeParamsB": 34.0, "totalParamsB": 34.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "qwen1532b", "name": "Qwen1.5-32B", "org": "Alibaba", "releaseDate": "2024-04-03", "openWeights": true, "capability": 30.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.3074 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.0155, "energyChatRange": [ 0.0155, 0.0155 ], "energyComplex": 0.361673, "energyComplexRange": [ 0.361673, 0.361673 ], "activeParamsB": 32.0, "totalParamsB": 32.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "mistralnemo", "name": "Mistral NeMo", "org": "Mistral AI", "releaseDate": "2024-07-18", "openWeights": true, "capability": 29.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.2989, "MATH level 5": 0.1083 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.013306, "energyChatRange": [ 0.013306, 0.013306 ], "energyComplex": 0.310477, "energyComplexRange": [ 0.310477, 0.310477 ], "activeParamsB": 12.0, "totalParamsB": 12.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "mixtral8x7b", "name": "Mixtral 8x7B", "org": "Mistral AI", "releaseDate": "2023-12-11", "openWeights": true, "capability": 29.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.2983, "MATH level 5": 0.0995 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.025004, "energyChatRange": [ 0.024389, 0.025619 ], "energyComplex": 0.583427, "energyComplexRange": [ 0.569082, 0.597772 ], "activeParamsB": [ 1.868, 7.472 ], "totalParamsB": 46.7, "sizeSource": "epoch_moe_estimated", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "isParetoFrontier": false }, { "id": "qwen1572b", "name": "Qwen1.5-72B", "org": "Alibaba", "releaseDate": "2024-02-04", "openWeights": true, "capability": 28.8, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.2882 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.079554, "energyChatRange": [ 0.079554, 0.079554 ], "energyComplex": 1.856263, "energyComplexRange": [ 1.856263, 1.856263 ], "activeParamsB": 72.0, "totalParamsB": 72.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "phi3medium14b", "name": "phi-3-medium 14B", "org": "Microsoft", "releaseDate": "2024-04-23", "openWeights": true, "capability": 27.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.2759, "MATH level 5": 0.1756 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.013526, "energyChatRange": [ 0.013526, 0.013526 ], "energyComplex": 0.315597, "energyComplexRange": [ 0.315597, 0.315597 ], "activeParamsB": 14.0, "totalParamsB": 14.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gemma29b", "name": "Gemma 2 9B", "org": "Google DeepMind", "releaseDate": "2024-06-24", "openWeights": true, "capability": 27.5, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.2746, "MATH level 5": 0.2101, "OTIS Mock AIME 2024-2025": 0.0056 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 0.013003, "energyChatRange": [ 0.013003, 0.013003 ], "energyComplex": 0.303412, "energyComplexRange": [ 0.303412, 0.303412 ], "activeParamsB": 9.24, "totalParamsB": 9.24, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gpt35turbo", "name": "GPT-3.5 Turbo", "org": "OpenAI", "releaseDate": "2024-01-25", "openWeights": false, "capability": 27.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "FrontierMath-Tiers-1-3-v2-Private": 0.0, "GPQA diamond": 0.2718, "MATH level 5": 0.1589, "Mystery Game Puzzles": 0.03, "OTIS Mock AIME 2024-2025": 0.0222 }, "reasoning": { "status": "hidden", "note": "Raw reasoning is never returned - only a summary - but it is billed as output tokens and counted in output_tokens_details.reasoning_tokens.", "source": "https://developers.openai.com/api/docs/guides/reasoning" }, "energyChat": 0.0573, "energyChatRange": [ 0.04633, 0.068271 ], "energyComplex": 1.337011, "energyComplexRange": [ 1.08103, 1.592993 ], "activeParamsB": [ 20, 70 ], "totalParamsB": 70.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "day1_estimate", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements. Active parameter count is estimated, so treat as a day-one figure.", "speed": 79.4, "isParetoFrontier": false }, { "id": "ministral8b", "name": "Ministral 8B", "org": "Mistral AI", "releaseDate": "2024-10-16", "openWeights": true, "capability": 27.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.2715, "MATH level 5": 0.1494 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.011647, "energyChatRange": [ 0.011647, 0.011647 ], "energyComplex": 0.271758, "energyComplexRange": [ 0.271758, 0.271758 ], "activeParamsB": 8.02, "totalParamsB": 8.02, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "speed": 42.4, "isParetoFrontier": false }, { "id": "llama318b", "name": "Llama 3.1-8B", "org": "Meta AI", "releaseDate": "2024-07-23", "openWeights": true, "capability": 27.0, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.2696, "MATH level 5": 0.2288, "OTIS Mock AIME 2024-2025": 0.0167 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.012867, "energyChatRange": [ 0.012867, 0.012867 ], "energyComplex": 0.300238, "energyComplexRange": [ 0.300238, 0.300238 ], "activeParamsB": 8.0, "totalParamsB": 8.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "measured": { "energy": 0.016692, "tokens": 398, "config": "B200 x1, batch 32", "task": "lm-arena-chat", "ratioToModel": 0.978, "source": "ML.ENERGY" }, "measurementDisagreement": false, "isParetoFrontier": false }, { "id": "llama270b", "name": "Llama 2-70B", "org": "Meta AI", "releaseDate": "2023-07-18", "openWeights": true, "capability": 26.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.2633, "MATH level 5": 0.0329, "OTIS Mock AIME 2024-2025": 0.0 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.078676, "energyChatRange": [ 0.078676, 0.078676 ], "energyComplex": 1.835784, "energyComplexRange": [ 1.835784, 1.835784 ], "activeParamsB": 70.0, "totalParamsB": 70.0, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "llama38b", "name": "Llama 3-8B", "org": "Meta AI", "releaseDate": "2024-04-18", "openWeights": true, "capability": 26.1, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.2607, "MATH level 5": 0.0613, "OTIS Mock AIME 2024-2025": 0.0194 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.012867, "energyChatRange": [ 0.012867, 0.012867 ], "energyComplex": 0.300238, "energyComplexRange": [ 0.300238, 0.300238 ], "activeParamsB": 8.0, "totalParamsB": 8.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "ministral3b", "name": "Ministral 3B", "org": "Mistral AI", "releaseDate": "2024-10-16", "openWeights": false, "capability": 25.3, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.2525, "MATH level 5": 0.1444 }, "reasoning": { "status": "undisclosed", "source": null, "note": "Closed model with no published position on whether reasoning text is returned." }, "energyChat": 0.011757, "energyChatRange": [ 0.011757, 0.011757 ], "energyComplex": 0.274321, "energyComplexRange": [ 0.274321, 0.274321 ], "activeParamsB": 3.32, "totalParamsB": 3.32, "sizeSource": "ecologits", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "speed": 35.0, "isParetoFrontier": false }, { "id": "deepseekllm67b", "name": "DeepSeek LLM 67B", "org": "DeepSeek", "releaseDate": "2023-11-29", "openWeights": true, "capability": 24.6, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.2462, "MATH level 5": 0.0639, "OTIS Mock AIME 2024-2025": 0.0083 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.07736, "energyChatRange": [ 0.07736, 0.07736 ], "energyComplex": 1.805067, "energyComplexRange": [ 1.805067, 1.805067 ], "activeParamsB": 67.0, "totalParamsB": 67.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "llama321b", "name": "Llama 3.2 1B", "org": "Meta AI", "releaseDate": "2024-09-24", "openWeights": true, "capability": 23.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.2393, "OTIS Mock AIME 2024-2025": 0.0056 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.012125, "energyChatRange": [ 0.012125, 0.012125 ], "energyComplex": 0.282908, "energyComplexRange": [ 0.282908, 0.282908 ], "activeParamsB": 1.23, "totalParamsB": 1.23, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gemma34b", "name": "Gemma 3 4B", "org": "Google DeepMind", "releaseDate": "2025-03-12", "openWeights": true, "capability": 23.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.2323, "OTIS Mock AIME 2024-2025": 0.075 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 0.012429, "energyChatRange": [ 0.012429, 0.012429 ], "energyComplex": 0.289999, "energyComplexRange": [ 0.289999, 0.289999 ], "activeParamsB": 4.0, "totalParamsB": 4.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "gemma31b", "name": "Gemma 3 1B", "org": "Google DeepMind", "releaseDate": "2025-03-12", "openWeights": true, "capability": 19.9, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.1995, "OTIS Mock AIME 2024-2025": 0.0111 }, "reasoning": { "status": "hidden", "note": "Full thoughts are generated internally and not returned; only summaries are, and only on request. Pricing counts the full thought tokens.", "source": "https://ai.google.dev/gemini-api/docs/thinking" }, "energyChat": 0.012099, "energyChatRange": [ 0.012099, 0.012099 ], "energyComplex": 0.282319, "energyComplexRange": [ 0.282319, 0.282319 ], "activeParamsB": 1.0, "totalParamsB": 1.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "yi34b", "name": "Yi-34B", "org": "01.AI", "releaseDate": "2023-11-22", "openWeights": true, "capability": 14.7, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "GPQA diamond": 0.1474, "MATH level 5": 0.0515 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.031439, "energyChatRange": [ 0.031439, 0.031439 ], "energyComplex": 0.733586, "energyComplexRange": [ 0.733586, 0.733586 ], "activeParamsB": 34.0, "totalParamsB": 34.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false }, { "id": "mistral7b", "name": "Mistral 7B", "org": "Mistral AI", "releaseDate": "2024-05-27", "openWeights": true, "capability": 13.2, "capabilitySource": "Epoch AI / GPQA diamond", "benchmarks": { "Chess Puzzles": 0.0, "GPQA diamond": 0.1323, "MATH level 5": 0.0368, "OTIS Mock AIME 2024-2025": 0.0028 }, "reasoning": { "status": "shown", "source": null, "note": "Open weights - the model can be run and read directly, so nothing it generates is hidden from you." }, "energyChat": 0.012758, "energyChatRange": [ 0.012758, 0.012758 ], "energyComplex": 0.297678, "energyComplexRange": [ 0.297678, 0.297678 ], "activeParamsB": 7.0, "totalParamsB": 7.0, "sizeSource": "epoch_dense", "architectureKnown": true, "energySource": "modeled", "energyNote": "EcoLogits formula, 7000-token answer, world-average grid, calibrated against ML.ENERGY measurements.", "isParetoFrontier": false } ] }