OpenVLM MMT-Bench - TU: leaderboard
The "TU" subtask of OpenVLM MMT-Bench. OpenCompass OpenVLM evaluation of MMT-Bench: multimodal tasks covering visual recognition, localization, OCR, counting, relation reasoning, visual math, and hallucination detection.
Metric: Score (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 207 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 (2025-04-14) | 57.5 | #250 |
| 2 | Qwen 2 VL 72B | 55.5 | #460 |
| 3 | Claude 3.5 Sonnet | 50.5 | #350 |
| 4 | InternVL3-38B | 49.6 | #405 |
| 5 | Grok 2 (1212) | 49.5 | #496 |
| 6 | InternVL3-8B | 48.7 | #620 |
| 7 | GPT-4.1 Mini | 48.4 | #355 |
| 8 | InternVL3-78B | 47.7 | #352 |
| 9 | InternVL3-14B | 47.7 | #505 |
| 10 | Pixtral-12B | 44.4 | #808 |
| 11 | Gemma 3 12B | 44.4 | #671 |
| 12 | Qwen 2.5 VL 7B | 43.4 | #643 |
| 13 | GPT-4.1 Nano | 42.5 | #731 |
| 14 | Qwen 2 VL 7B | 41.6 | #750 |
| 15 | MiniCPM-V-2.6 | 41.5 | #836 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmt-bench-tu · How It Works · Data refreshed daily, snapshot 2026-10-20.