MMMU-Pro
Canonical page on the main site: chinaaihub.com/benchmarks/mmmu-pro
Description
Multimodal, multi-discipline understanding benchmark with college-level questions requiring reasoning.
Evaluations
| benchmark | model | score | metric | date | source_type | source_url |
|---|---|---|---|---|---|---|
| MMMU-Pro | kimi-k3 | 81.6 (83.4 with tools) | accuracy | 2026-07 | vendor_reported | https://github.com/MoonshotAI/Kimi-K3 |
| MMMU-Pro | kimi-k2.5 | 78.5 | accuracy | — | vendor_reported | https://github.com/MoonshotAI/Kimi-K2.5 |
Limitations
All scores are vendor-reported and not independently verified. With-tools and without-tools results are not directly comparable.
Last Verified
2026-09-20
Sources
| source_name | source_url | source_type | last_verified | confidence |
|---|---|---|---|---|
| Kimi K3 GitHub README | https://github.com/MoonshotAI/Kimi-K3 | official | 2026-09-20 | high |
| Kimi K2.5 GitHub README | https://github.com/MoonshotAI/Kimi-K2.5 | official | 2026-09-20 | high |