Xiaomi · model
MiMo-V2.5 benchmark scores
As of 2026-10-07, MiMo-V2.5 (Xiaomi) has measured scores on 13 benchmarks and estimated scores on 67 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 82.5% | estimated ± 4.2 pp, high confidence |
| AA-HLE | 24.0% | estimated ± 7.4 pp, medium confidence |
| AA-SciCode | 48.3% | estimated ± 3.8 pp, high confidence |
| Artificial Analysis Intelligence Index | 20.9% | estimated ± 7.9 pp, medium confidence |
| CritPt | 3.6% | estimated ± 6.6 pp, medium confidence |
| GDPval-AA | 32.6% | estimated ± 8.3 pp, low confidence |
| GPQA Diamond | 85.8% | estimated ± 1.7 pp, medium confidence |
| HLE | 37.3% | estimated ± 9.8 pp, low confidence |
| LiveCodeBench v6 | 87.9% | estimated ± 2.1 pp, high confidence |
| AA-Omniscience Accuracy | 24.9% | estimated ± 7.7 pp, medium confidence |
| SciCode | 39.7% | estimated ± 2.9 pp, low confidence |
| Terminal-Bench 2.1 | 78.3% | estimated ± 6.9 pp, medium confidence |
| BrowseComp | 63.6% | estimated ± 5.7 pp, medium confidence |
| VITA-Bench | 39.6% | estimated ± 10.7 pp, low confidence |
| AA Coding Index | 54.3% | estimated ± 5.8 pp, medium confidence |
| AA-MMMU-Pro | 75.9% | estimated ± 4.9 pp, medium confidence |
| APEX-Agents-AA | 31.8% | estimated ± 10.0 pp, low confidence |
| DeepSearchQA | 80.5% | estimated ± 10.2 pp, low confidence |
| SWE-bench Verified | 79.4% | estimated ± 2.5 pp, high confidence |
| BFCL v4 | 62.3% | estimated ± 8.6 pp, low confidence |
| CyberGym | 41.1% | estimated ± 10.0 pp, low confidence |
| JobBench | 31.8% | estimated ± 8.4 pp, low confidence |
| SWE-bench Pro | 56.1% | measured |
| τ³-bench results | 68.9% | estimated ± 8.9 pp, medium confidence |
| WideResearch | 70.4% | estimated ± 5.9 pp, low confidence |
| AA Agentic Index | 19.7% | estimated ± 5.1 pp, medium confidence |
| GPQA | 86.9% | estimated ± 1.6 pp, high confidence |
| Gert Labs | 46.9% | measured |
| ApprenticeBench | 0.0% | estimated ± 3.1 pp, low confidence |
| ARC-AGI-1 | 72.6% | estimated ± 5.9 pp, medium confidence |
| ARC-AGI-2 | 24.6% | estimated ± 12.6 pp, low confidence |
| CursorBench 3.1 | 52.0% | estimated ± 5.6 pp, low confidence |
| CursorBench 3.2 | 52.3% | estimated ± 3.6 pp, high confidence |
| Terminal-Bench 3.0 | 6.1% | estimated ± 6.9 pp, low confidence |
| FrontierCode 1.1 Main | 25.1% | estimated ± 5.2 pp, low confidence |
| FrontierSWE v2 | 8.2% | estimated ± 9.0 pp, medium confidence |
| OSWorld-Verified | 58.0% | estimated ± 5.9 pp, medium confidence |
| Vals GPQA Diamond | 81.6% | measured |
| Vals LiveCodeBench | 81.5% | measured |
| Vals MMLU-Pro | 82.9% | measured |
| Vals SWE-bench | 71.0% | measured |
| Terminal-Bench 2.1 (Vals AI) | 60.7% | measured |
| VulcanBench v3 | 81.6% | estimated ± 5.6 pp, low confidence |
| AA Terminal-Bench 2.1 | 77.5% | estimated ± 4.4 pp, high confidence |
| AA Terminal-Bench 4.0 | 14.5% | estimated ± 14.0 pp, low confidence |
| CursorBench 4.0 | 32.5% | estimated ± 3.6 pp, medium confidence |
| DeepSWE | 56.3% | estimated ± 7.2 pp, low confidence |
| HLE w/o tools | 26.5% | estimated ± 5.1 pp, medium confidence |
| OSWorld 2.0 | 0.0% | estimated ± 4.5 pp, low confidence |
| PostTrainBench v1.1 | 20.5% | estimated ± 2.7 pp, high confidence |
| CharXiv | 81.0% | measured |
| CharXiv w/o tools | 82.1% | estimated ± 1.8 pp, medium confidence |
| Claw-Eval | 62.3% | measured |
| MathVision | 86.8% | estimated ± 2.6 pp, high confidence |
| MCP Atlas | 57.0% | estimated ± 11.2 pp, low confidence |
| MCP-Tasks | 77.5% | estimated ± 6.4 pp, low confidence |
| MMMU-Pro | 77.9% | measured |
| NL2Repo | 41.1% | estimated ± 4.7 pp, high confidence |
| ScreenSpot Pro | 73.6% | estimated ± 7.1 pp, medium confidence |
| Terminal-Bench 2.0 | 65.8% | measured |
| Toolathlon | 52.3% | estimated ± 6.9 pp, medium confidence |
| VideoMMMU | 84.8% | estimated ± 1.0 pp, high confidence |
| ERQA | 66.5% | estimated ± 3.7 pp, high confidence |
| LiveCodeBench Pro | 80.6% | estimated ± 6.0 pp, low confidence |
| MedXpertQA (MM) | 70.0% | estimated ± 2.4 pp, high confidence |
| React Native Evals | 67.1% | estimated ± 2.4 pp, low confidence |
| ResearchClawBench | 16.9% | measured |
| ARC-AGI-3 | 0.0% | estimated ± 9.2 pp, low confidence |
| FrontierCode 1.1 Extended | 17.0% | estimated ± 2.4 pp, low confidence |
| HealthBench Professional | 50.1% | estimated ± 1.3 pp, low confidence |
| HLE-Verified | 0.0% | estimated ± 4.6 pp, low confidence |
| LABBench2 | 79.1% | estimated ± 1.8 pp, low confidence |
| MMMU | 84.4% | estimated ± 1.1 pp, medium confidence |
| OpenHarmony Bench | 51.5% | estimated ± 3.2 pp, high confidence |
| SimpleVQA | 67.1% | estimated ± 7.8 pp, low confidence |
| MMMU-Pro w/ Python | 79.3% | estimated ± 0.6 pp, high confidence |
| MathVision w/ Python | 76.8% | estimated ± 0.9 pp, low confidence |
| RealWorldQA | 85.4% | estimated ± 1.1 pp, medium confidence |
| Video-MME (with subtitle) | 87.7% | measured |
| OmniDocBench 1.5 | 91.7% | estimated ± 5.9 pp, low confidence |