Google DeepMind·released 2023-05-17
Scored on 5 benchmarks. 0% of its scores are independently reproduced — the rest are self-reported or unverified.
Every score, ranked, colored by evidence status.
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| HellaSwag | 76accuracy (%) | self-reported· optimized | 2023-05-17 |
| TriviaQA | 75.2accuracy (%) | self-reported· optimized | 2023-05-17 |
| Winogrande | 55.8accuracy (%) | self-reported· optimized | 2023-05-17 |
| ARC AI2 | 46.13accuracy (%) | unverified· optimized | 2023-05-17 |
| OpenBookQA | 41.6accuracy (%) | self-reported· optimized | 2023-05-17 |