GPT-2 (1.5B) is an open-weights model from OpenAI, released November 5, 2019.
| Domain | Benchmark | |
|---|---|---|
| Language understanding | PIQABest score 89% | |
WinoGrandeBest score 89% | ||
BoolQBest score 91% | ||
HellaSwagBest score 95% | ||
| Science | OpenBookQABest score 88% | |
ARC AI2Best score 95% | ||
| Software engineering | METR Time HorizonsBest score 17 hr 25 min | 3 sec |