ChemBench: leaderboard
2,700+ chemistry questions across 9 sub-topics (organic, inorganic, analytical, physical, materials science, toxicity). Published in Nature Chemistry. 40 models evaluated.
Metric: Overall Score (%). Source: chembench.lamalab.org. Saturation forecast: Estimated already saturated. 40 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.7 Sonnet (20250219) | 65.83 | #201 |
| 2 | O1 Preview | 64 | #193 |
| 3 | GPT-4o | 61 | #343 |
| 4 | Llama 3.1 405B Instruct | 58 | #457 |
| 5 | Mistral Large 2 (Jul) | 57 | #484 |
| 6 | Llama 3.1 70B Instruct | 53 | #561 |
| 7 | Qwen 2.5 32B | 52.87 | #546 |
| 8 | Llama 3 70B Instruct | 52 | #634 |
| 9 | Gemma 2 9B (IT) | 48 | #788 |
| 10 | Llama 3.1 8B Instruct | 47 | #1035 |
| 11 | GPT-3.5 Turbo | 47 | #865 |
| 12 | Phi-3-medium-4k-instruct | 47 | #952 |
| 13 | Claude 2 | 47 | #761 |
| 14 | Llama 3 8B Instruct | 46 | #1129 |
| 15 | Command-R+ | 45 | #891 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=chembench · How It Works · Data refreshed daily, snapshot 2026-10-20.