Math-VR: leaderboard
Mathematical visual reasoning benchmark for VLMs, unified models, and LLMs, reporting answer correctness and process scores on text and multimodal questions.
Metric: Overall Answer Correctness (self-reported). Source: benchmarklist.com. Status: saturation imminent. 30 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 VL 235B A22B (Thinking) | 66.8 |
| 2 | Qwen 3 VL 235B A22B Instruct | 65 |
| 3 | Gemini 2.5 Pro | 64.7 |
| 4 | Gemini 2.5 Flash | 60.5 |
| 5 | O3 | 59.3 |
| 6 | GPT-5 (Thinking) | 58.1 |
| 7 | Gemini 2.5 Flash (Thinking) | 52.3 |
| 8 | GLM-4.5V | 49.6 |
| 9 | GPT-4.1 Mini | 33.3 |
| 10 | Claude Sonnet 4 | 28.1 |
| 11 | GPT-4.1 | 26 |
| 12 | Gemini 2.0 Flash | 20.6 |
| 13 | GPT-4.1 Nano | 9.1 |
| 14 | GPT-4o | 4.3 |
Interactive version: theaggregate.ai/benchmark?slug=math-vr · How It Works · Data refreshed daily, snapshot 2026-09-05.