SciEval - Overall: leaderboard
SciEval multimodal scientific-reasoning leaderboard — overall accuracy across exam-style science questions.
Metric: Overall (%). Source: huggingface.co. Saturation forecast: Around March 2027. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 65.59 | #98 |
| 2 | O3 | 65.38 | #124 |
| 3 | Gemini 3 Pro | 63.9 | #80 |
| 4 | Qwen 3 Max | 61.49 | #208 |
| 5 | Qwen 3 VL 235B A22B | 60.58 | #238 |
| 6 | GPT-5.1 | 57.73 | #135 |
| 7 | Gemini 2.5 Pro | 57.54 | #152 |
| 8 | Kimi K2 | 56.91 | #243 |
| 9 | Intern-S1 | 56.73 | #285 |
| 10 | Claude Sonnet 4.5 | 56.27 | #142 |
| 11 | Llama 4 Maverick | 53.74 | #464 |
| 12 | Ling-flash-2.0 | 52.49 | #550 |
| 13 | GPT-4o | 52.38 | #343 |
| 14 | Gemini 2.5 Flash | 50.85 | #245 |
| 15 | Grok 2 (1212) | 46.01 | #496 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=scieval-overall · How It Works · Data refreshed daily, snapshot 2026-10-20.