multimodal benchmark
MMStar Leaderboard
MMStar is a multimodal benchmark with 1,500 human-selected challenge samples for evaluating 6 core capabilities and 18 detailed axes.
Updated Sep 5, 2026
Models25
Model coverage25
MetricScore
EvidenceB
MMStar Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAC | Score83.30% | Percentile100.00% | Participants25 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score82.90% | Percentile95.83% | Participants25 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score81.90% | Percentile91.67% | Participants25 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score81.40% | Percentile87.50% | Participants25 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Score81.00% | Percentile83.33% | Participants25 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Score79.40% | Percentile79.17% | Participants25 | EvidenceC | Evaluated |
| Rank07 | ModelAC | Score78.70% | Percentile75.00% | Participants25 | EvidenceC | Evaluated |
| Rank08 | ModelAC | Score78.40% | Percentile70.83% | Participants25 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score77.70% | Percentile66.67% | Participants25 | EvidenceC | Evaluated |
| Rank10 | ModelAC | Score75.50% | Percentile62.50% | Participants25 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score75.30% | Percentile58.33% | Participants25 | EvidenceC | Evaluated |
| Rank12 | ModelLA | Score74.90% | Percentile54.17% | Participants25 | EvidenceC | Evaluated |
| Rank13 | ModelAC | Score73.20% | Percentile50.00% | Participants25 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score72.10% | Percentile45.83% | Participants25 | EvidenceC | Evaluated |
| Rank15 | ModelAC | Score70.90% | Percentile41.67% | Participants25 | EvidenceC | Evaluated |
| Rank16 | ModelAC | Score70.80% | Percentile37.50% | Participants25 | EvidenceC | Evaluated |
| Rank17 | ModelAC | Score69.80% | Percentile33.33% | Participants25 | EvidenceC | Evaluated |
| Rank18 | ModelAC | Score69.50% | Percentile29.17% | Participants25 | EvidenceC | Evaluated |
| Rank19 | ModelAC | Score64.00% | Percentile25.00% | Participants25 | EvidenceC | Evaluated |
| Rank20 | ModelAC | Score63.90% | Percentile20.83% | Participants25 | EvidenceC | Evaluated |
| Rank21 | ModelLA | Score63.30% | Percentile16.67% | Participants25 | EvidenceC | Evaluated |
| Rank22 | ModelDE | Score61.30% | Percentile12.50% | Participants25 | EvidenceC | Evaluated |
| Rank23 | ModelDE | Score57.00% | Percentile8.33% | Participants25 | EvidenceC | Evaluated |
| Rank24 | ModelCO | Score51.80% | Percentile4.17% | Participants25 | EvidenceC | Evaluated |
| Rank25 | ModelDE | Score45.90% | Percentile0.00% | Participants25 | EvidenceC | Evaluated |
MMStar Highlights
The leading models and scores on this benchmark.
MMStar Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for MMStar
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is MMStar?
What MMStar measures and how its scores work.
MMStar is a multimodal benchmark comprising 1,500 challenge samples selected by humans.
It evaluates 6 core capabilities and 18 detailed axes using the Score metric, reported as a ratio.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- MMStar
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about MMStar.
Which model scores highest on MMStar?
Qwen3.6 Plus is currently ranked first with 83.30%.
What are the top three models on MMStar?
The current leaders are Qwen3.6 Plus (83.30%), Qwen3.5-122B-A10B (82.90%), and Qwen3.5-35B-A3B (81.90%).
Which MMStar model has the lowest official input price?
Qwen2.5-Omni-7B has the lowest matched official input price at $0.10 input / $0.40 output per 1M tokens.
Which models are fastest among MMStar results?
The fastest matched records are Qwen3 VL 4B Thinking (8.77 tok/s via DeepInfra).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.