OpenVLM MathVista - GEO: leaderboard
OpenCompass OpenVLM MathVista geometry reasoning subcategory score.
Metric: Accuracy (%). Source: huggingface.co. Source: huggingface.co. Saturation forecast: Estimated already saturated. 284 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 95 | #98 |
| 2 | GPT-5 Mini (2025-08-07) | 90.8 | #172 |
| 3 | GPT-5 Nano | 90.4 | #425 |
| 4 | Doubao-1.5-Pro | 89.1 | #287 |
| 5 | Gemini 2.5 Pro | 88.3 | #152 |
| 6 | InternVL3-78B | 83.3 | #352 |
| 7 | Gemini 2.0 Flash | 81.6 | #342 |
| 8 | Gemini 1.5 Pro (002) | 80.3 | #349 |
| 9 | URSA-8B | 79.5 | #811 |
| 10 | Gemma 3 27B | 78.2 | #607 |
| 11 | InternVL3-38B | 77.8 | #405 |
| 12 | InternVL3-14B | 77 | #505 |
| 13 | GPT-4.5 | 76.2 | #274 |
| 14 | InternVL3-8B | 75.7 | #620 |
| 15 | Grok 2 (1212) | 74.1 | #496 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mathvista-geo · How It Works · Data refreshed daily, snapshot 2026-10-20.