Meta · model
Muse Spark benchmark scores
As of 2026-11-21, Muse Spark (Meta) has measured scores on 36 benchmarks and estimated scores on 112 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 88.4% | measured |
| AA-HLE | 40.7% | measured |
| AA-SciCode | 48.7% | estimated ± 3.4 pp, high confidence |
| Artificial Analysis Intelligence Index | 31.3% | measured |
| CritPt | 11.3% | measured |
| GDPval-AA | 24.9% | measured |
| GPQA Diamond | 89.5% | measured |
| HLE | 50.4% | measured |
| IFBench | 74.2% | estimated ± 6.6 pp, medium confidence |
| AA-LCR | 78.0% | measured |
| LiveCodeBench v6 | 86.6% | estimated ± 2.7 pp, high confidence |
| AA-Omniscience Accuracy | 49.6% | measured |
| SciCode | 46.4% | estimated ± 3.6 pp, high confidence |
| BrowseComp | 60.1% | estimated ± 5.2 pp, medium confidence |
| HLE w/ tools | 50.0% | estimated ± 4.9 pp, medium confidence |
| IFEval | 93.0% | estimated ± 2.8 pp, high confidence |
| LongBench v2 | 64.0% | estimated ± 5.5 pp, medium confidence |
| VITA-Bench | 41.0% | estimated ± 10.7 pp, low confidence |
| τ²-bench results | 91.5% | measured |
| AA Coding Index | 58.6% | measured |
| AA-MMMU-Pro | 80.5% | measured |
| APEX-Agents | 18.0% | estimated ± 9.4 pp, low confidence |
| APEX-Agents-AA | 14.0% | estimated ± 6.0 pp, low confidence |
| BioMysteryBench (human-difficult) | 44.7% | estimated ± 6.8 pp, low confidence |
| DeepSearchQA | 74.8% | measured |
| SWE-bench Verified | 77.4% | measured |
| BFCL v4 | 73.4% | estimated ± 2.7 pp, high confidence |
| CyberGym | 43.5% | measured |
| JobBench | 31.2% | estimated ± 8.6 pp, low confidence |
| SWE-bench Pro | 52.4% | measured |
| WideResearch | 70.1% | estimated ± 1.6 pp, medium confidence |
| AA Agentic Index | 19.1% | estimated ± 3.1 pp, high confidence |
| MMLU-Pro | 85.7% | estimated ± 2.5 pp, low confidence |
| AA-IFBench | 75.9% | measured |
| GPQA | 89.5% | estimated ± 0.0 pp, medium confidence |
| MMLU | 90.5% | estimated ± 0.9 pp, medium confidence |
| Vals GPQA Diamond | 89.6% | measured |
| Vals LiveCodeBench | 79.7% | estimated ± 4.7 pp, medium confidence |
| Vals MMLU-Pro | 87.3% | measured |
| MMMLU | 89.7% | estimated ± 0.7 pp, medium confidence |
| MMMU | 86.4% | estimated ± 0.8 pp, medium confidence |
| AA MMLU-Pro | 88.8% | estimated ± 0.7 pp, high confidence |
| Gert Labs | 54.8% | estimated ± 5.5 pp, medium confidence |
| AA EnterpriseOps-Gym | 43.2% | estimated ± 2.3 pp, medium confidence |
| AA Tau3 Banking | 29.9% | estimated ± 4.1 pp, high confidence |
| ApprenticeBench | 8.4% | estimated ± 10.1 pp, low confidence |
| ARC-AGI-1 | 82.0% | estimated ± 1.7 pp, high confidence |
| ARC-AGI-2 | 42.5% | measured |
| CursorBench 3.1 | 47.9% | estimated ± 4.8 pp, low confidence |
| CursorBench 3.2 | 45.4% | estimated ± 3.6 pp, medium confidence |
| FrontierCode 1.1 Main | 22.9% | estimated ± 4.2 pp, low confidence |
| FrontierSWE v2 | 2.5% | estimated ± 8.3 pp, low confidence |
| MazeBench | 0.4% | estimated ± 5.5 pp, low confidence |
| OfficeQA Pro | 60.4% | estimated ± 6.8 pp, medium confidence |
| OSWorld-Verified | 71.0% | estimated ± 3.8 pp, medium confidence |
| ParseBench-1 | 61.9% | estimated ± 5.2 pp, low confidence |
| Terminal-Bench 4.1 | 2.2% | estimated ± 9.2 pp, low confidence |
| Terminal-Bench Hard | 45.5% | measured |
| Vals SWE-bench | 74.4% | measured |
| Terminal-Bench 2.1 (Vals AI) | 57.4% | estimated ± 5.6 pp, medium confidence |
| CWE-Bench-AA | 32.4% | estimated ± 10.1 pp, low confidence |
| DeepsecBench-AA | 1.2% | estimated ± 8.9 pp, low confidence |
| HealthBench Professional (length-adjusted, Artificial Analysis) | 55.4% | estimated ± 4.3 pp, low confidence |
| HealthBench Professional (raw, Artificial Analysis) | 57.2% | estimated ± 0.8 pp, low confidence |
| AA ITBench | 37.3% | estimated ± 5.1 pp, low confidence |
| Bug Hunt Bench | 7.4% | estimated ± 9.0 pp, low confidence |
| Chartography (tools) | 83.7% | estimated ± 4.1 pp, low confidence |
| CursorBench 4.0 | 29.8% | estimated ± 3.3 pp, medium confidence |
| CWE-bench v1 | 34.3% | estimated ± 6.4 pp, low confidence |
| DeepSWE | 30.0% | estimated ± 6.4 pp, low confidence |
| DRACO | 72.4% | estimated ± 2.0 pp, medium confidence |
| FrontierCode 2.0 | 19.3% | estimated ± 6.7 pp, low confidence |
| HLE w/o tools | 42.8% | measured |
| OSWorld 2.1 Offline | 2.4% | estimated ± 1.9 pp, low confidence |
| PostTrainBench v1.1 | 19.3% | estimated ± 3.7 pp, medium confidence |
| Toolathlon-Verified | 51.0% | estimated ± 3.1 pp, medium confidence |
| VulcanBench Frontier v4 | 75.1% | estimated ± 8.3 pp, low confidence |
| GDP.pdf | 17.0% | estimated ± 4.6 pp, medium confidence |
| ARC-AGI-3 | 4.8% | estimated ± 7.2 pp, medium confidence |
| ARC-AGI-3 Provider Adapter | 2.7% | estimated ± 7.3 pp, medium confidence |
| MKA | 21.1% | estimated ± 6.4 pp, low confidence |
| Vibe Code Bench | 19.7% | measured |
| Hermes Bench | 55.7% | estimated ± 2.9 pp, high confidence |
| Hermes Index | 22.8% | estimated ± 8.7 pp, medium confidence |
| ParseBench-2 | 72.7% | estimated ± 5.0 pp, medium confidence |
| CharXiv | 86.4% | measured |
| CharXiv w/o tools | 84.6% | estimated ± 1.2 pp, high confidence |
| ExploitGym | 1.5% | estimated ± 9.3 pp, low confidence |
| AI2D_TEST | 93.5% | estimated ± 0.9 pp, low confidence |
| C-Eval | 94.9% | estimated ± 1.0 pp, low confidence |
| Claw-Eval | 63.8% | measured |
| CountBench | 96.1% | estimated ± 2.2 pp, medium confidence |
| DynaMath | 87.4% | estimated ± 0.8 pp, low confidence |
| ERQA | 64.7% | measured |
| MathVision | 100.0% | estimated ± 2.2 pp, low confidence |
| MCP Atlas | 62.8% | estimated ± 10.7 pp, low confidence |
| MCP-Tasks | 80.4% | estimated ± 6.4 pp, low confidence |
| MLVU (M-Avg) | 87.8% | estimated ± 1.5 pp, high confidence |
| MMLU-Redux | 97.2% | estimated ± 0.8 pp, low confidence |
| MMMU-Pro | 80.4% | measured |
| NL2Repo | 42.1% | estimated ± 4.7 pp, high confidence |
| OmniDocBench 1.5 | 89.8% | estimated ± 2.0 pp, medium confidence |
| PostTrainBench v1 | 19.0% | estimated ± 3.1 pp, medium confidence |
| QwenClawBench | 55.9% | estimated ± 6.0 pp, low confidence |
| RealWorldQA | 87.6% | estimated ± 1.4 pp, medium confidence |
| ScreenSpot Pro | 84.1% | measured |
| SimpleVQA | 71.3% | measured |
| SuperGPQA | 89.6% | estimated ± 7.2 pp, medium confidence |
| Terminal-Bench 2.0 | 59.0% | measured |
| Toolathlon | 43.1% | estimated ± 6.3 pp, medium confidence |
| V* | 96.1% | estimated ± 6.5 pp, low confidence |
| Video-MME (w/o subtitle) | 85.7% | estimated ± 1.3 pp, high confidence |
| Video-MME (with subtitle) | 89.3% | estimated ± 2.9 pp, high confidence |
| VideoMMMU | 86.4% | estimated ± 1.5 pp, medium confidence |
| We-Math | 83.8% | estimated ± 2.1 pp, medium confidence |
| AA LiveCodeBench | 88.6% | estimated ± 2.6 pp, high confidence |
| HealthBench Hard | 42.8% | measured |
| LiveCodeBench Pro | 80.0% | measured |
| MedXpertQA (MM) | 78.4% | measured |
| MedXpertQA (Text) | 52.6% | measured |
| MMLU-Pro (Arcee) | 86.7% | estimated ± 3.4 pp, low confidence |
| React Native Evals | 78.5% | estimated ± 2.7 pp, medium confidence |
| ResearchClawBench | 17.7% | estimated ± 1.3 pp, high confidence |
| SWE-Rebench | 60.6% | estimated ± 3.6 pp, medium confidence |
| PostTrainBench | 16.7% | estimated ± 3.7 pp, medium confidence |
| BioMysteryBench (human-solvable) | 84.4% | estimated ± 0.6 pp, low confidence |
| FrontierCode 1.1 Extended | 53.3% | estimated ± 3.0 pp, low confidence |
| HealthBench (raw) | 60.3% | estimated ± 0.7 pp, low confidence |
| HealthBench (length-adjusted) | 60.1% | estimated ± 0.4 pp, low confidence |
| HealthBench Professional | 52.1% | estimated ± 4.2 pp, medium confidence |
| HealthBench Professional (raw) | 74.3% | estimated ± 1.7 pp, low confidence |
| HLE-Verified | 15.4% | estimated ± 2.0 pp, low confidence |
| LABBench2 | 81.6% | estimated ± 1.8 pp, low confidence |
| ScarfBench | 25.9% | estimated ± 9.1 pp, low confidence |
| STEMBench | 2.6% | estimated ± 0.2 pp, medium confidence |
| SimpleQA | 32.8% | estimated ± 7.8 pp, low confidence |
| LiveCodeBench | 77.7% | estimated ± 14.6 pp, low confidence |
| BabyVision w/ Python | 88.9% | estimated ± 3.8 pp, low confidence |
| OpenHarmony Bench v1.0 | 51.4% | estimated ± 2.8 pp, high confidence |
| BabyVision | 66.2% | estimated ± 9.1 pp, low confidence |
| MMVU | 78.6% | estimated ± 2.7 pp, medium confidence |
| PerceptionBench | 68.3% | estimated ± 2.7 pp, medium confidence |
| ZeroBench | 33.0% | measured |
| MRCRv2 | 85.5% | estimated ± 11.3 pp, low confidence |
| MMLongBench-Doc | 61.3% | estimated ± 2.6 pp, low confidence |
| OCRBench V2 | 71.7% | estimated ± 3.3 pp, high confidence |
| BioBasicsBench | 34.0% | estimated ± 2.8 pp, medium confidence |
| MMMU-Pro w/ Python | 83.2% | estimated ± 2.0 pp, high confidence |