SciEval - Chemistry: leaderboard
SciEval multimodal scientific-reasoning leaderboard — chemistry-discipline accuracy.
Metric: Chemistry (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 83.08 | #80 |
| 2 | O3 | 81.77 | #124 |
| 3 | GPT-5 | 81.62 | #98 |
| 4 | Gemini 2.5 Pro | 78.29 | #152 |
| 5 | Qwen 3 VL 235B A22B | 77.39 | #238 |
| 6 | DeepSeek R1 | 77.28 | #251 |
| 7 | Kimi K2 | 77.04 | #243 |
| 8 | GPT-5.1 | 76.45 | #135 |
| 9 | Qwen 3 Max | 76.38 | #208 |
| 10 | Intern-S1 | 75.28 | #285 |
| 11 | Gemini 2.5 Flash | 75.12 | #245 |
| 12 | GPT-4o | 72.37 | #343 |
| 13 | GLM-4.5V | 71.38 | #346 |
| 14 | Claude Sonnet 4.5 | 71.27 | #142 |
| 15 | Llama 4 Maverick | 68.06 | #464 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=scieval-chemistry · How It Works · Data refreshed daily, snapshot 2026-10-20.