VideoMMMU
We show this table for reference; we do not rank on it.
A video extension of MMMU-style multimodal reasoning over expert questions grounded in temporal media.
Benchmark score on VideoMMMU — October 10, 2026
We compile the VideoMMMU rows from provider self-reports and secondary reports. Qwen3.8 Max leads the table at 88.7%, followed by Gemini 3 Pro (87.6%) and dots3-note Preview (86.8%). We do not use these results to rank models overall.
Qwen3.8 Max
Alibaba
Gemini 3 Pro
dots3-note Preview
Dots Studio
12 modelsMultimodal & GroundedCurrentDisplay onlyUpdated October 10, 2026
Benchmark score table (12 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Qwen3.8 MaxAlibaba | 88.7% | Not reported | Open |
| 2 | Gemini 3 ProGoogle | 87.6% | Not reported | Closed |
| 3 | dots3-note PreviewDots Studio | 86.8% | Not reported | Open |
| 4 | Kimi K2.5Moonshot AI | 86.6% | Not reported | Open |
| 5 | GPT-5.2OpenAI | 85.9% | Not reported | Closed |
| 6 | Qwen3.7 PlusAlibaba | 85.4% | Not reported | Closed |
| 7 | Qwen3.5 397BAlibaba | 84.7% | Not reported | Open |
| 8 | MiniMax M3MiniMax | 84.6% | Not reported | Open |
| 9 | Qwen3.6-27BAlibaba | 84.4% | Not reported | Open |
| 10 | Claude Opus 4.5Anthropic | 84.4% | Not reported | Closed |
| 11 | Qwen3.6 PlusAlibaba | 84.0% | Not reported | Closed |
| 12 | Qwen3.6-35B-A3BAlibaba | 83.7% | Not reported | Open |
Among the reported VideoMMMU rows, Qwen3.8 Max is first at 88.7%. The third row is 1.9 points behind. The broader top-10 range is 4.3 points, so many of the published results sit in a relatively narrow band.
12 models have been evaluated on VideoMMMU. The benchmark falls in the Multimodal & Grounded category. VideoMMMU is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About VideoMMMU
Year
2026
Tasks
Video-grounded expert reasoning
Format
Video + text reasoning
Difficulty
Frontier multimodal video reasoning
VideoMMMU tests whether multimodal reasoning skills extend from static images into temporal video understanding. It is useful for evaluating long-form visual reasoning rather than static scene recognition.
Freshness and provenance
Version
VideoMMMU 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does VideoMMMU measure?
A video extension of MMMU-style multimodal reasoning over expert questions grounded in temporal media.
Which model scores highest on VideoMMMU?
Qwen3.8 Max by Alibaba currently leads with a score of 88.7% on VideoMMMU.
How many models are evaluated on VideoMMMU?
12 AI models have published results on VideoMMMU in the BenchLM catalog.
Compare top models on VideoMMMU
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 5,500+ readers.
One email each week. Unsubscribe anytime.