Same standard of work. About 6.5× the energy.
GPT-5.6 Luna scores higher than o3 on Epoch's Capabilities Index, 156.67 against 146.86, across 58 benchmarks covering coding, agents, long context and writing, while using about 6.5× less energy per answer. In blind head-to-head voting, people rate them 1431 and 1410 respectively.
| GPT-5.6 Luna | o3 | |
|---|---|---|
| Energy per answer | 2.61 Wh | 17.0 Wh |
| Carbon released | 1.2 g CO₂ | 7.8 g CO₂ |
| Same carbon as driving | 5 metres (16 feet) | 31 metres (103 feet) |
| Energy rank | 170 of 281 | 239 of 281 |
| Capability (58 benchmarks) | 156.7 | 146.9 |
| Human vote rating | 1431 | 1410 |
| Votes cast | 41,317 | 58,901 |
| Confidence | Day-one estimate | Inferred |
| Released | 2026-07-09 | 2025-04-16 |
Energy is calculated the same way for every model on this site, from public data and calibrated against laboratory measurement, accurate to roughly a factor of two to three, which is why we only ever state a difference of 5× or more. Capability is Epoch AI's Capabilities Index, one score combining 58 separate benchmarks; the vote rating comes from blind human comparisons. The full method is here.