Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Multimodal & Grounded benchmark
Multimodal Multi-disciplinary Video Understanding. Every model the catalog carries a published MMVU value for, ranked by that value.
A benchmark for evaluating multimodal models on video understanding tasks across multiple disciplines, emphasizing temporal reasoning and comprehension over video content. 8 models in the ModelScale catalog have a published value.
Each point is a model with a published MMVU value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
8 models with a published MMVU value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Video reasoning benchmark |
|---|---|---|---|
| 1 | Alibaba | 82.4 | |
| 2 | Z.AI | 80.5 | |
| 3 | Moonshot AI | 80.4 |
What the benchmark measures and how its value is published, from the source’s own record.
MMVU (Multimodal Multi-disciplinary Video Understanding) is a multimodal & grounded benchmark published by Kimi K2.5 benchmark release surface. A benchmark for evaluating multimodal models on video understanding tasks across multiple disciplines, emphasizing temporal reasoning and comprehension over video content. Its evaluation set is described by the source as video understanding. The source rates its difficulty as multi-disciplinary multimodal video reasoning.
Each model's result is published by Kimi K2.5 benchmark release surface as video reasoning benchmark. ModelScale prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 8 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
All leaderboards · Multimodal & Grounded capability leaderboard
| 4 | DSdots3-note Preview | Dots Studio | 79.9 |
| 5 | Alibaba | 74.7 |
| 6 | Alibaba | 73.3 |
| 7 | Alibaba | 72.3 |
| 8 | LiquidAI | 52.9 |
Evidence key: Observed
Rows are ordered by the value Kimi K2.5 benchmark release surface published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.