OpenVLM MMT-Bench - Handwritten Retrieval: leaderboard
Metric: Score (%). Source: huggingface.co. 207 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 1.5 Pro | 95 |
| 2 | GPT-4.1 (2025-04-14) | 90 |
| 3 | GPT-4.1 Mini | 75 |
| 4 | Gemini 1.5 Flash | 75 |
| 5 | InternVL3-78B | 60 |
| 6 | InternVL3-38B | 55 |
| 7 | Grok 2 (1212) | 55 |
| 8 | InternVL3-8B | 35 |
| 9 | Qwen 2 VL 7B | 35 |
| 10 | GPT-4.1 Nano | 35 |
| 11 | Claude 3.5 Sonnet | 35 |
| 12 | Gemma 3 4B | 35 |
| 13 | Pixtral-12B | 30 |
| 14 | Claude 3.7 Sonnet | 30 |
| 15 | Llama 3.2 90B Vision Instruct | 30 |
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmt-bench-handwritten-retrieval · How It Works · Data refreshed daily, snapshot 2026-09-05.