These panels summarize public benchmark signals from Xiaomi MiMo source pages. They are not a replacement for the original tables; they are a faster reading layer on top of them.
MiMo-V2-Flash is publicly framed as a reasoning-forward model family, with strong results on math and knowledge-heavy tasks.
| Benchmark | Context | Value |
|---|---|---|
| MMLU-Pro | MiMo-V2-Flash | 84.9 |
| GPQA-Diamond | MiMo-V2-Flash | 83.7 |
| AIME 2025 | MiMo-V2-Flash | 94.1 |
Public benchmark tables emphasize coding, SWE-Bench, and terminal-style agent tasks as part of the MiMo-V2-Flash story.
| Benchmark | Context | Value |
|---|---|---|
| LiveCodeBench-v6 | MiMo-V2-Flash | 80.6 |
| SWE-Bench Verified | MiMo-V2-Flash | 73.4 |
| Terminal-Bench 2.0 | MiMo-V2-Flash | 38.5 |
The earlier MiMo release remains useful for understanding the public lineage and the reasoning-first positioning of the wider model family.
| Benchmark | Context | Value |
|---|---|---|
| MATH500 | MiMo-7B-RL | 95.8 |
| AIME 2024 | MiMo-7B-RL | 68.2 |
| LiveCodeBench v6 | MiMo-7B-RL | 49.3 |