Open LMM Reasoning - MathVision - analytic geometry: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 100 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 (2025-04-14) | 58.3 |
| 2 | Gemini 2.0 Flash | 51.2 |
| 3 | GPT-4o ChatGPT | 51.2 |
| 4 | GPT-4.1 Mini | 48.8 |
| 5 | Doubao-1.5-Pro | 48.8 |
| 6 | QVQ-72B-Preview | 46.4 |
| 7 | Claude 3.7 Sonnet | 44 |
| 8 | Claude 3.5 Sonnet (20241022) | 41.7 |
| 9 | Gemini 1.5 Pro (002) | 41.7 |
| 10 | GPT-4.1 Nano | 39.3 |
| 11 | InternVL2.5-78B-CoT | 39.3 |
| 12 | InternVL3-8B | 38.1 |
| 13 | InternVL3-38B | 38.1 |
| 14 | Gemma 3 12B | 38.1 |
| 15 | VLAA-Thinker-Qwen2.5VL-3B | 36.9 |
Interactive version: theaggregate.ai/benchmark?slug=open-lmm-reasoning-mathvision-analytic-geometry · How It Works · Data refreshed daily, snapshot 2026-09-05.