NatureBench — leaderboard
Metric: Surpass-SOTA (%). Source: frontisai.github.io. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.7 | 17.78 |
| 2 | Gemini 3.5 Flash | 15.56 |
| 3 | GLM-5.2 | 15.56 |
| 4 | GPT-5.5 | 14.44 |
| 5 | Claude Opus 4.6 | 12.22 |
| 6 | MiniMax-M3 | 11.11 |
| 7 | Qwen 3.7 Max | 10 |
| 8 | GPT-5.4 | 8.89 |
| 9 | Kimi K2.6 | 8.89 |
| 10 | GLM-5.1 | 7.78 |
| 11 | DeepSeek V4 Pro | 4.44 |
| 12 | MiniMax-M2.7 | 1.11 |
Interactive version: theaggregate.ai/benchmark?slug=naturebench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.