SciEval - Astronomy: leaderboard
SciEval multimodal scientific-reasoning leaderboard — astronomy-discipline accuracy.
Metric: Astronomy (%). Source: huggingface.co. Saturation forecast: Around December 2027. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 Max | 75.64 | #208 |
| 2 | Ling-flash-2.0 | 48.12 | #550 |
| 3 | GPT-5 | 44.57 | #98 |
| 4 | Gemini 2.5 Pro | 43.4 | #152 |
| 5 | O3 | 42.82 | #124 |
| 6 | Gemini 3 Pro | 42.23 | #80 |
| 7 | GPT-5.1 | 42.02 | #135 |
| 8 | Kimi K2 | 35.57 | #243 |
| 9 | Claude Sonnet 4.5 | 34.23 | #142 |
| 10 | Grok 2 (1212) | 33.51 | #496 |
| 11 | Gemini 2.5 Flash | 31.94 | #245 |
| 12 | Llama 4 Maverick | 31.91 | #464 |
| 13 | GPT-4o | 30.73 | #343 |
| 14 | Qwen 3 VL 235B A22B | 29.84 | #238 |
| 15 | Intern-S1 | 29.55 | #285 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=scieval-astronomy · How It Works · Data refreshed daily, snapshot 2026-10-20.