- Best of Humanitys Last Exam 34.0 / with tools 48.0 48
- GDPVal-AA (Elo) 1581
- τ³-bench 72.9
- Claw-Eval (pass^3) 63.8
- SWE-Bench Pro 57.2
- SWE-bench Verified 78.9
- Terminal-Bench 2.0 68.4
- MiMo Coding Bench 73.7
MiMo-V2.5-Pro
MiMo-V2.5-Pro ranks first on CompassArena.
73
Benchmarks with a score
1
Ranked first, out of five or more models
3
In the top three
Xiaomi Corp Apr 2026
MiMo-V2.5-Pro
Open weights
154.3
#65 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- Claw-Eval 80.6% #2 /24
- Claw-Eval Pass^3 63.8% #5 /13
- Harvey LAB (Legal Agent Benchmark) 2.08% #48 /80
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CompassArena | 1,062 | #1 of 10 | Official leaderboard | |
| LMArena (Elo) | 1,467 | #43 of 335 | Official leaderboard | |
| StudyArena Text | 1,270 | #14 of 85 | Official leaderboard | |
| Longform Creative Writing (EQ-Bench) | 73.7 | #30 of 119 | Official leaderboard | |
| BullshitBench V2 | 62% | #34 of 129 | Official leaderboard | |
| BullshitBench | 45.5% | #32 of 107 | Official leaderboard | |
| Humanity's Last Exam | 48% | #49 of 154 | Lab model card | |
| Creative Writing v3 | 71.4 | #46 of 120 | Official leaderboard | |
| Judgemark v4 | 57.9 | #21 of 40 | Official leaderboard | |
| DesignArena ASCII Art | 1,151 | #45 of 77 | Official leaderboard | |
| EQ-Bench 4 | 1,198 | #19 of 32 | Official leaderboard | |
| Extended NYT Connections | 34.4% | #80 of 103 | Official leaderboard | |
| OpenAI MRCR v2 | 18.6% | #40 of 47 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DTBench | 84.5% | #72 of 180 | Official leaderboard | |
| ProofBench | 24% | #27 of 59 | Official leaderboard | |
| ProofBench v1.1 | 22% | #40 of 52 | Official leaderboard | |
| ASI-Bench | 23.3 | #11 of 11 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SWE-bench Verified | 78.9% | #24 of 155 | Lab model card | |
| DesignArena Game Dev | 1,265 | #37 of 143 | Official leaderboard | |
| DesignArena Website | 1,264 | #40 of 151 | Official leaderboard | |
| DesignArena Code | 1,262 | #43 of 143 | Official leaderboard | |
| DesignArena Data Visualization | 1,258 | #38 of 118 | Official leaderboard | |
| DesignArena 3D Design | 1,235 | #44 of 136 | Official leaderboard | |
| DesignArena UI Component | 1,244 | #55 of 139 | Official leaderboard | |
| AutoLab | 0.45 | #4 of 10 | Official leaderboard | |
| MiMo Coding Bench | 73.7% | #2 of 5 | Official leaderboard | |
| DesignArena SVG | 1,183 | #46 of 103 | Official leaderboard | |
| WebDev Arena | 1,475 | #49 of 105 | arena.ai, via Epoch AI | |
| SWE-Bench Pro | 57.2% | #40 of 81 | Lab model card | |
| FeatureBench | 47.8% | #8 of 16 | Official leaderboard | |
| FeatureBench Lite | 47.8% | #8 of 16 | Official leaderboard | |
| Code Arena | 1,479 | #65 of 128 | Official leaderboard | |
| EvoCode-Bench | 17.3% | #8 of 13 | Official leaderboard | |
| SWE-rebench | 46.5% | #8 of 13 | Official leaderboard | |
| Vibe Code Bench | 34.1% | #76 of 120 | Official leaderboard | |
| ALE-Bench | 1,254 | #80 of 126 | Official leaderboard | |
| Code Migration | 21.6% | #50 of 78 | Official leaderboard | |
| AndroidBench | 60.8% | #23 of 32 | Official leaderboard | |
| RoadmapBench | 13.9% | #8 of 11 | Official leaderboard | |
| GameCraft-Bench | 24.1% | #12 of 15 | Official leaderboard | |
| MiMo Code Bench | 40.4% | #8 of 8 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Claw-Eval | 80.6% | #2 of 24 | Official leaderboard | |
| PinchBench 2.0 | 87.5% | #9 of 57 | Official leaderboard | |
| Terminal-Bench 2.0 | 68.4% | #10 of 63 | Lab model card | |
| Clocktower Radio | 1,684 | #6 of 23 | Official leaderboard | |
| ClawProBench | 63.3% | #13 of 49 | Official leaderboard | |
| tau3-bench | 72.9% | #4 of 14 | Lab model card | |
| SnakeBench | 27 | #57 of 188 | Official leaderboard | |
| Claw-Eval Pass^3 | 63.8% | #5 of 13 | Lab model card | |
| LiveClawBench | 70.6% | #7 of 17 | Official leaderboard | |
| PACT | 1,508 | #17 of 26 | Official leaderboard | |
| WildClawBench | 43% | #22 of 31 | Official leaderboard | |
| Harbor-Index | 2.40% | #9 of 9 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| TaxEval v2 | 73.8% | #45 of 129 | Official leaderboard | |
| MedScribe | 83.7% | #40 of 97 | Official leaderboard | |
| GDPval-AA | 1,581 | #7 of 16 | Lab model card | |
| CorpFin v2 | 61.4% | #54 of 119 | Official leaderboard | |
| FoodTruck Bench | $22,388 | #31 of 59 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 2.08% | #48 of 80 | Official leaderboard | |
| Excel Modeling Benchmark | 55.2% | #49 of 76 | Official leaderboard | |
| Vals Index v2 | 41% | #44 of 65 | Official leaderboard | |
| PhysicianBench | 16.7% | #9 of 13 | Official leaderboard | |
| LegalBench | 77.1% | #102 of 138 | Official leaderboard | |
| Vals Finance Agent v2 | 41.5% | #61 of 80 | Official leaderboard | |
| Legal Research Bench | 15.9% | #74 of 93 | Official leaderboard | |
| Tax Agent Bench | 11.5% | #68 of 84 | Official leaderboard | |
| MedCode | 32.5% | #82 of 96 | Official leaderboard | |
| Vals Index v2.1 | 33.8% | #44 of 49 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| nocot-bench (NCRI) | 77.1 | #68 of 185 | Official leaderboard | |
| MiMo Cyber Bench | 0% | #5 of 5 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Conceptual Reasoning Index | 44.2 | #89 of 151 | Official leaderboard | |
| LMCA | 29.5% | #91 of 147 | Official leaderboard | |
| ACCoRD | 58.2 | #95 of 152 | Official leaderboard |
Its lab's numbers put it at 154.3, number 65 of 404
The Self-Reported Index fits one score to the 6 capability results Xiaomi Corp reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 151.3 to 157.4. How we compute it.
- SWE-bench Verified 78.9
- tau3-bench 72.9
- Terminal-Bench 2.0 68.4
- Claw-Eval Pass^3 63.8
- SWE-Bench Pro 57.2
- Humanity's Last Exam 48
Its lab's numbers next to 2 leaderboards
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
| Benchmark | Lab says | Board says | Gap |
|---|---|---|---|
| Claw-Eval Pass^3 | 63.8% Claw-Eval (pass^3) | 63.8% Leaderboard | Same |
| MiMo Coding Bench | 73.7% MiMo Coding Bench | 73.7% Leaderboard | Same |
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.