LLM Stats (MathVision): leaderboard
3,040 math problems with visual contexts (geometric figures, graphs, diagrams) sourced from real math competitions and textbooks across 16 difficulty levels.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Around January 2027. 38 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.7 Plus | 90.3 | #92 |
| 2 | Qwen 3.6 Plus | 88 | #120 |
| 3 | Qwen 3.5 122B A10B | 86.2 | #173 |
| 4 | Qwen 3.5 27B | 86 | #187 |
| 5 | Gemma 4 31B | 85.6 | #200 |
| 6 | Kimi K2.5 | 84.2 | #144 |
| 7 | Qwen 3.5 35B A3B | 83.9 | #256 |
| 8 | Gemma 4 26B A4B | 82.4 | #358 |
| 9 | Seed 1.8 | 81.3 | #139 |
| 10 | Gemma 4 12B | 79.7 | #535 |
| 11 | Nemotron 3.5 Super VL (BF16) | 78.16 | #165 |
| 12 | Seed 2.0 Mini | 78.1 | #276 |
| 13 | Qwen 3 VL 235B A22B (Thinking) | 74.6 | #238 (Qwen 3 VL 235B A22B) |
| 14 | Step3 VL 10B | 70.8 | #485 |
| 15 | DiffusionGemma 26B-A4B | 70.5 | #698 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-mathvision · How It Works · Data refreshed daily, snapshot 2026-10-20.