SciEval - Earth Sciences: leaderboard
SciEval multimodal scientific-reasoning leaderboard — earth-sciences-discipline accuracy.
Metric: Earth Sciences (%). Source: huggingface.co. Saturation forecast: Around January 2027. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2 | 77.77 | #243 |
| 2 | GPT-5 | 74.43 | #98 |
| 3 | O3 | 74.15 | #124 |
| 4 | GPT-5.1 | 73.88 | #135 |
| 5 | Gemini 3 Pro | 72.49 | #80 |
| 6 | Gemini 2.5 Pro | 70.52 | #152 |
| 7 | Ling-flash-2.0 | 69.55 | #550 |
| 8 | Intern-S1 | 68.52 | #285 |
| 9 | Gemini 2.5 Flash | 68.21 | #245 |
| 10 | Qwen 3 VL 235B A22B | 67.81 | #238 |
| 11 | Qwen 3 Max | 67.38 | #208 |
| 12 | Claude Sonnet 4.5 | 64.66 | #142 |
| 13 | DeepSeek R1 | 63.14 | #251 |
| 14 | GPT-4o | 63.08 | #343 |
| 15 | Llama 4 Maverick | 59.82 | #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-earth-sciences · How It Works · Data refreshed daily, snapshot 2026-10-20.