OpenRouter GPQA Diamond: leaderboard
OpenRouter's own GPQA Diamond run: 198 graduate-level science questions executed by its native benchmark harness against the production endpoints it serves, so the score reflects the deployed model rather than a vendor-reported figure.
Metric: Accuracy (%). Source: openrouter.ai. Status: saturation imminent. 134 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 94.4 |
| 2 | GPT-6 | 94.4 |
| 3 | Gemini 3.7 Flash | 94.3 |
| 4 | GPT-5.5 | 93.8 |
| 5 | GPT-5.6 Pro Sol | 93.8 |
| 6 | Grok 4.6 | 93.3 |
| 7 | Gemini 3.5 Flash | 92.8 |
| 8 | Gemini 3.6 Flash | 92.8 |
| 9 | GPT-5.6 Sol | 91.9 |
| 10 | Kimi K3 | 91.9 |
| 11 | MiniMax-M3 | 90.6 |
| 12 | GPT-5.4 | 90.3 |
| 13 | GPT-5.6 Pro Luna | 90.3 |
| 14 | Claude Fable 5.1 | 90.1 |
| 15 | Claude Opus 4.8 | 89.7 |
Interactive version: theaggregate.ai/benchmark?slug=openrouter-gpqa-diamond · How It Works · Data refreshed daily, snapshot 2026-09-20.