OpenVLM OCRBench - Handwritten Mathematical Expression Recognition: leaderboard
Metric: Correct Answers (%). Source: huggingface.co. 285 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Moonshot v1 8K | 89 |
| 2 | InternVL3-78B | 85 |
| 3 | InternVL3-8B | 83 |
| 4 | InternVL3-14B | 79 |
| 5 | InternVL2.5-78B | 78 |
| 6 | InternVL2.5-2B | 76 |
| 7 | Qwen 2 VL 72B | 73 |
| 8 | Aquila-VL-2B | 64 |
| 9 | InternVL2-8B | 63 |
| 10 | Qwen 2 VL 7B | 59 |
| 11 | Qwen 2 VL 2B | 39 |
| 12 | Gemini 2.5 Pro | 38 |
| 13 | Gemini 2.0 Flash | 38 |
| 14 | GPT-4.5 | 34 |
| 15 | GPT-4.1 (2025-04-14) | 31 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-ocrbench-handwritten-mathematical-expression-recognition · How It Works · Data refreshed daily, snapshot 2026-09-19.