GeneBench-Pro Not Externally Reviewed
Science · 2026-06-30
GeneBench-Pro not externally reviewed stratum reports pass rates on the 47 problems in the 129-problem GeneBench-Pro suite that were not externally reviewed.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.6 Sol Pro | 36.6 |
| GPT-5.6 Sol | 34.1 |
| GPT-5.6 Terra Pro | 30.6 |
| GPT-5.6 Luna Pro | 29.4 |
| GPT-5.6 Terra | 28.7 |
| GPT-5.5 Pro | 24.7 |
| GPT-5.6 Luna | 20.9 |
| Opus 4.8 | 20.4 |
| GPT-5.4 Pro | 17.9 |
| GPT-5.5 | 16.6 |
| Gemini 3.5 Flash | 13.0 |
| GPT-5.4 | 12.6 |
| GPT-5.2 Pro | 10.6 |
| GPT-5.2 | 7.0 |
| GLM-5.2 | 6.0 |
| Kimi K2.6 | 4.9 |
| Qwen3.7-Max | 4.0 |
| Gemini 3.1 Pro Preview | 3.4 |
| Kimi K2.7 Code | 3.2 |
| DeepSeek-V4-Flash | 3.1 |
| Grok 4.3 | 3.0 |
| Qwen3.7-Plus | 3.0 |
| DeepSeek-V4-Pro | 2.4 |
| MiMo-V2.5-Pro | 1.9 |
| MiniMax M3 | 1.3 |
| MiMo-V2.5 | 1.3 |
| GLM-5.1 | 1.1 |
| MiniMax M2.7 | 0.9 |
| Hy3 preview | 0.9 |