text-davinci-001: benchmark results
InstructGPT completion model snapshot: the original GPT-3 base supervised fine-tuned on human demonstrations. Provider: OpenAI. Released 2022-01-27. Access: API.
Aggregate Rating 1278 ± 33, rank #1512 of 1656 rated models, from 16 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Epoch AI - Lambada | 86.4 | Score | 96 |
| Epoch AI - Science Qa | 74.04 | Score | 73.9 |
| OpenBookQA | 65.4 | Accuracy (%) | 73.8 |
| WinoGrande | 77.7 | Accuracy (%) | 66.2 |
| PIQA | 82.3 | Accuracy (%) | 65 |
| HellaSwag | 79.3 | Accuracy (%) | 61.8 |
| ARC Challenge (AI2) | 53.2 | Accuracy (%) | 45.4 |
| BoolQ | 77.5 | Accuracy (%) | 41.6 |
| TriviaQA | 71.2 | Accuracy (%) | 26.3 |
| AlpacaEval 2.0 | 9.03 | LC Win Rate (%) | 18.9 |
| Epoch AI - Common Sense Qa 2 | 52.9 | Score | 16.7 |
| MMLU | 43.9 | Accuracy (%) | 16.4 |
Interactive version: theaggregate.ai/model?slug=text-davinci-001 · How It Works · Data refreshed daily, snapshot 2026-10-20.