GPT-4 (Mar 2023) is a closed-weights model from OpenAI, released March 14, 2023. It has an ECI score of 126 and ranks 173rd of 249 tracked models. At the time of its release, it ranked 1st of 5 tracked models. At $30/Mtok for input tokens and $60/Mtok for output tokens, it is more expensive per token than similarly-capable models.
| Domain | Benchmark | ||
|---|---|---|---|
ECI scoreBest score 166 | 126 | 110 | |
| Ranking | #173 | #216 | |
| Language understanding | WinoGrandeBest score 89.2% | ||
HellaSwagBest score 95.3% | |||
| Mathematics | GSM8KBest score 94.5% | ||
OTIS Mock AIME 2024-2025Best score 100.0% | |||
| Science | GPQA DiamondBest score 95.8% | ||
| Software engineering | METR Time HorizonsBest score 1044.8 | 5.4 | — |
| World knowledge | MMLUBest score 88.1% |