STEM Reasoning on GPQA-Diamond (Pass@1)
91.9Pass@1Gemini 3-pro
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini 3-proDataset=GPQA-Diamond2026.03 | 91.9 | |
| Gemini 2.5-proDataset=GPQA-Diamond2026.03 | 86.4 | |
| GPT-5 HighDataset=GPQA-Diamond2026.03 | 85.7 | |
| Seed1.8Dataset=GPQA-Diamond2026.03 | 83.8 | |
| Claude Sonnet-4.5Dataset=GPQA-Diamond2026.03 | 83.4 | |
| Gemini 3-proDataset=PHYBench2026.03 | 59 | |
| Gemini 3-proDataset=BIOBench2026.03 | 51.9 | |
| Gemini 2.5-proDataset=PHYBench2026.03 | 48 | |
| GPT-5 HighDataset=BIOBench2026.03 | 48 | |
| Claude Sonnet-4.5Dataset=BIOBench2026.03 | 44.6 | |
| Seed1.8Dataset=BIOBench2026.03 | 42.3 | |
| Gemini 2.5-proDataset=BIOBench2026.03 | 41.5 | |
| Seed1.8Dataset=PHYBench2026.03 | 41 | |
| GPT-5 HighDataset=PHYBench2026.03 | 40 | |
| Claude Sonnet-4.5Dataset=PHYBench2026.03 | 31 |