OpenVLM MMBench V1.1 CN - FP-S: leaderboard
OpenCompass OpenVLM MMBench V1.1 Chinese split accuracy for single-instance fine-grained perception questions.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around January 2027. 284 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 93.3 | #152 |
| 2 | InternVL3-78B | 93.3 | #352 |
| 3 | Qwen 2 VL 72B | 92.8 | #460 |
| 4 | InternVL2.5-78B | 92.8 | #431 |
| 5 | InternVL3-38B | 92.3 | #405 |
| 6 | Grok 2 (1212) | 91.2 | #496 |
| 7 | Qwen 2.5 VL 7B | 89.6 | #643 |
| 8 | Claude 3.5 Sonnet (20241022) | 87.9 | #291 |
| 9 | Gemini 1.5 Pro (002) | 87.9 | #349 |
| 10 | GPT-4.1 (2025-04-14) | 87.9 | #250 |
| 11 | InternVL3-14B | 87.9 | #505 |
| 12 | InternVL3-8B | 86.8 | #620 |
| 13 | Gemini 1.5 Flash (002) | 86.8 | #583 |
| 14 | Qwen 2 VL 7B | 86.1 | #750 |
| 15 | Llama 3.2 90B Vision Instruct | 85.4 | #622 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmbench-v1-1-cn-fp-s · How It Works · Data refreshed daily, snapshot 2026-10-20.