GPT-2 is the scrappy ancestor of modern language models — it can string together coherent sentences and mimic writing styles, but it loses the thread quickly and hallucinates freely. It has no instruction-following ability; it simply continues whatever text you give it. Think of it as a pattern-matching engine that learned from the internet, useful for understanding how language models work at a foundational level.
| Benchmark | Score | Type | Recorded |
|---|---|---|---|
| GPQA Diamond | 1.1 | accuracy | 26d ago |
| BBH | 2.7 | accuracy | 26d ago |
| MuSR | 15.3 | accuracy | 26d ago |
| MMLU-Pro | 1.8 | accuracy | 26d ago |
| IFEval | 17.9 | accuracy | 26d ago |
| MATH | 0.2 | accuracy | 26d ago |