Benchmark
MMStar Benchmark
What is MMStar?
MMStar is an elite vision-indispensable multimodal benchmark comprising 1,500 challenge samples meticulously selected by humans to evaluate 6 core capabilities…
- Released
- 2024-04-09
- Modality
- multimodal
- Evaluates
- multimodal, reasoning, general, vision
- Openness
- restricted
- Importer
- LLM Stats
- Reported scores
- 24
Source provenance
- Original evidence https://api.zeroeval.com/leaderboard/benchmarks/mmstar?top_n=500
MMStar paper, code and dataset
MMStar results and reported scores
LLM Stats
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| DeepSeek VL2 | DeepSeek | 0.613 | 2024-12-13 | evidence |
| DeepSeek VL2 Small | DeepSeek | 0.57 | 2024-12-13 | evidence |
| DeepSeek VL2 Tiny | DeepSeek | 0.459 | 2024-12-13 | evidence |
| LFM2.5-VL-3B | Liquid AI | 0.633 | 2026-08-12 | evidence |
| North Micro Vision Instruct | Cohere | 0.518 | 2026-08-12 | evidence |
| Qwen2.5 VL 32B Instruct | Qwen | 0.695 | 2025-02-28 | evidence |
| Qwen2.5 VL 72B Instruct | Qwen | 0.708 | 2025-01-26 | evidence |
| Qwen2.5 VL 7B Instruct | Qwen | 0.639 | 2025-01-26 | evidence |
| Qwen2.5-Omni-7B | Qwen | 0.64 | 2025-03-27 | evidence |
| Qwen3 VL 235B A22B Instruct | Qwen | 0.784 | 2025-09-22 | evidence |
| Qwen3 VL 235B A22B Thinking | Qwen | 0.787 | 2025-09-22 | evidence |
| Qwen3 VL 30B A3B Instruct | Qwen | 0.721 | 2025-09-22 | evidence |
| Qwen3 VL 30B A3B Thinking | Qwen | 0.755 | 2025-09-22 | evidence |
| Qwen3 VL 32B Instruct | Qwen | 0.777 | 2025-09-22 | evidence |
| Qwen3 VL 32B Thinking | Qwen | 0.794 | 2025-09-22 | evidence |
| Qwen3 VL 4B Instruct | Qwen | 0.698 | 2025-09-22 | evidence |
| Qwen3 VL 4B Thinking | Qwen | 0.732 | 2025-09-22 | evidence |
| Qwen3 VL 8B Instruct | Qwen | 0.709 | 2025-09-22 | evidence |
| Qwen3 VL 8B Thinking | Qwen | 0.753 | 2025-09-22 | evidence |
| Qwen3.5-122B-A10B | Qwen | 0.829 | 2026-02-24 | evidence |
| Qwen3.5-27B | Qwen | 0.81 | 2026-02-24 | evidence |
| Qwen3.5-35B-A3B | Qwen | 0.819 | 2026-02-24 | evidence |
| Qwen3.6 Plus | Qwen | 0.833 | 2026-04-02 | evidence |
| Qwen3.6-27B | Qwen | 0.814 | 2026-04-21 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.