multimodal benchmark
MVBench Leaderboard
A comprehensive multi-modal video understanding benchmark covering 20 challenging video tasks that require temporal understanding beyond single-frame analysis. Tasks span from perception to cognition, including action recognition, temporal reasoning, spatial reasoning, object interaction, scene transition, and counterfactual inference. Uses a novel static-to-dynamic method to systematically generate video tasks from existing annotations.
Updated Aug 17, 2026
Models17
Model coverage17
MetricScore
EvidenceB
MVBench Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAC | Score76.6% | Percentile100.0% | Participants17 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score75.5% | Percentile93.8% | Participants17 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score74.8% | Percentile87.5% | Participants17 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score74.6% | Percentile81.3% | Participants17 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Score74.6% | Percentile75.0% | Participants17 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Score73.6% | Percentile68.8% | Participants17 | EvidenceC | Evaluated |
| Rank07 | ModelAC | Score73.2% | Percentile62.5% | Participants17 | EvidenceC | Evaluated |
| Rank08 | ModelAC | Score72.8% | Percentile56.3% | Participants17 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score72.3% | Percentile50.0% | Participants17 | EvidenceC | Evaluated |
| Rank10 | ModelAC | Score72.0% | Percentile43.8% | Participants17 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score70.4% | Percentile37.5% | Participants17 | EvidenceC | Evaluated |
| Rank12 | ModelAC | Score70.3% | Percentile31.3% | Participants17 | EvidenceC | Evaluated |
| Rank13 | ModelAC | Score69.6% | Percentile25.0% | Participants17 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score69.3% | Percentile18.8% | Participants17 | EvidenceC | Evaluated |
| Rank15 | ModelAC | Score69.0% | Percentile12.5% | Participants17 | EvidenceC | Evaluated |
| Rank16 | ModelAC | Score68.9% | Percentile6.3% | Participants17 | EvidenceC | Evaluated |
| Rank17 | ModelAC | Score68.7% | Percentile0.0% | Participants17 | EvidenceC | Evaluated |
MVBench Highlights
The leading models and scores on this benchmark.
MVBench Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for MVBench
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis mvbench AI model leaderboard uses descending score in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
- 01ACQwen3.5-122B-A10BAlibaba Cloud / Qwen Team
- Score
- 76.6%
- Price
- $0.40 input / $3.2 output per 1M tokens
Strengths
- Ranks #1 of 17 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures MVBench, not total model capability
- 02ACQwen3.6-27BAlibaba Cloud / Qwen Team
- Score
- 75.5%
- Price
- $0.60 input / $3.6 output per 1M tokens
- Speed
- Up to 6.1 tok/s via Novita
Strengths
- Ranks #2 of 17 compared models
- 94th percentile on this benchmark
- C evidence result
Considerations
- This result measures MVBench, not total model capability
- 03ACQwen3.5-35B-A3BAlibaba Cloud / Qwen Team
- Score
- 74.8%
- Price
- $0.25 input / $2.0 output per 1M tokens
Strengths
- Ranks #3 of 17 compared models
- 88th percentile on this benchmark
- C evidence result
Considerations
- This result measures MVBench, not total model capability
- 04ACQwen3.5-27BAlibaba Cloud / Qwen Team
- Score
- 74.6%
- Price
- $0.30 input / $2.4 output per 1M tokens
- Speed
- Up to 6.7 tok/s via Novita
Strengths
- Ranks #4 of 17 compared models
- 81th percentile on this benchmark
- C evidence result
Considerations
- This result measures MVBench, not total model capability
- 05ACQwen3.6-35B-A3BAlibaba Cloud / Qwen Team
- Score
- 74.6%
- Price
- $0.25 input / $1.5 output per 1M tokens
Strengths
- Ranks #5 of 17 compared models
- 75th percentile on this benchmark
- C evidence result
Considerations
- This result measures MVBench, not total model capability
Selection summary
Best AI Models for MVBench
Qwen3.5-122B-A10B currently leads MVBench with 76.6%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.
What is MVBench?
What MVBench measures and how its scores work.
A comprehensive multi-modal video understanding benchmark covering 20 challenging video tasks that require temporal understanding beyond single-frame analysis. Tasks span from perception to cognition, including action recognition, temporal reasoning, spatial reasoning, object interaction, scene transition, and counterfactual inference. Uses a novel static-to-dynamic method to systematically generate video tasks from existing annotations.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- MVBench
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- mvbench|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about MVBench.
Which model scores highest on MVBench?
Qwen3.5-122B-A10B is currently ranked first with 76.6%.
What does MVBench measure?
A comprehensive multi-modal video understanding benchmark covering 20 challenging video tasks that require temporal understanding beyond single-frame analysis. Tasks span from perception to cognition, including action recognition, temporal reasoning, spatial reasoning, object interaction, scene transition, and counterfactual inference. Uses a novel static-to-dynamic method to systematically generate video tasks from existing annotations.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
17 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.