- Personal Superintelligence Index (internal composite) 54.3
- Harvey LAB-AA 10
- Vals Finance Agent V2 56.6
- HealthBench Pro 60.5
- AutomationBench-AA 39.1
- GDP.pdf 30.8
- JobBench 70.5
- OSWorld V2.1 (partial; binary 34.3) 72.6
- MemoryBench 85.6
- Terminal-Bench 4.0-AA 52
- DeepSWE 1.1 78.8
- SWE-Marathon (version not stated) 53.1
- ProgramBench
- Artificial Analysis Index 52
- FocusBench (chart label) 44.9
Muse Wonder
Muse Wonder ranks first on 5 of the 158 benchmarks with five or more models scored.
160
Benchmarks with a score
5
Ranked first, out of five or more models
30
In the top three
Meta Oct 2026 Muse Wonder
Closed weights
161.7
#15 of 256
Epoch Capabilities Index
Self-Reported Index 163.9 #16 of 404
- Terminal-Bench 4.0 47% #10 /40
- HealthBench Professional 60.5% #16 /26
- CritPt 69.7% #8 /11
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 | |
|---|---|---|---|---|
| LMArena (Elo) | 1,511 | #3 of 335 | Official leaderboard | |
| Creative Writing v3 | 81.1 | #7 of 120 | Official leaderboard | |
| EQ-Bench 4 | 1,325 | #3 of 32 | Official leaderboard | |
| AA-LCR Extended | 74.4% | #5 of 49 | Official leaderboard | |
| Extended NYT Connections | 93.4% | #14 of 103 | Official leaderboard | |
| StudyArena Text | 1,274 | #13 of 85 | Official leaderboard | |
| Lech Mazur Creative Writing | 1.90 | #12 of 50 | Official leaderboard | |
| BullshitBench V2 | 57.6% | #40 of 129 | Official leaderboard | |
| HLE-Diamond | 40.6% | #11 of 20 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 99.4% | #19 of 212 | Epoch AI run | |
| DTBench | 95.2% | #23 of 180 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 82.1% | #19 of 92 | Epoch AI run | |
| FrontierMath Tier 4 v2 | 70.7% | #18 of 74 | Epoch AI run | |
| ProofBench v1.1 | 84% | #14 of 52 | Official leaderboard | |
| MysteryMechanism | 27.5% | #16 of 31 | Official leaderboard | |
| CritPt | 69.7% | #8 of 11 | Official leaderboard | |
| CritPt v2 | 69.7% | #8 of 11 | Official leaderboard | |
| Prophet Arena | 85.3 | #18 of 24 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DesignArena Website | 1,380 | #1 of 151 | Official leaderboard | |
| DesignArena Code | 1,394 | #2 of 143 | Official leaderboard | |
| DesignArena Game Dev | 1,416 | #4 of 143 | Official leaderboard | |
| DeepSWE | 78.8% | #2 of 56 | Lab model card | |
| DesignArena UI Component | 1,368 | #5 of 139 | Official leaderboard | |
| DesignArena 3D Design | 1,434 | #6 of 136 | Official leaderboard | |
| Code Migration | 62.5% | #6 of 78 | Official leaderboard | |
| DesignArena Image to HTML | 1,264 | #6 of 51 | Official leaderboard | |
| Snorkel Agentic Coding 2.0 | 43.6% | #2 of 16 | Official leaderboard | |
| DesignArena Data Visualization | 1,290 | #16 of 118 | Official leaderboard | |
| Code Arena | 1,649 | #19 of 128 | Official leaderboard | |
| SWE-Marathon v1.1 | 52.5% | #3 of 20 | Official leaderboard | |
| ProgramBench | 5.50% | #4 of 25 | Official leaderboard | |
| ProgramBench tests passed | 86.5% | #4 of 25 | Official leaderboard | |
| DesignArena Image to Frontend | 1,256 | #5 of 31 | Official leaderboard | |
| Code Migration v1.1 | 72.4% | #6 of 30 | Official leaderboard | |
| ProgramBench almost solved | 50.5% | #5 of 25 | Official leaderboard | |
| Vibe Code Bench | 83.1% | #25 of 120 | Official leaderboard | |
| Vibe Code Bench 1-100 | 26.6% | #6 of 26 | Official leaderboard | |
| SWE-Marathon | 53.1% | #3 of 11 | Lab model card | |
| SWE-Bench Pro V2 Hard | 90.2% | #5 of 18 | Official leaderboard | |
| SWE Atlas v1.1 | 64.8% | #9 of 28 | Official leaderboard | |
| CursorBench | 50.4% | #7 of 21 | cursor.com, via Epoch AI | |
| IOI v2 (Vals) | 66.9% | #16 of 45 | Official leaderboard | |
| SWE-bench CLI | 11.4% | #6 of 16 | Official leaderboard | |
| CursorBench 4.0 | 50.4% | #8 of 21 | Official leaderboard | |
| SWE-Bench Pro V2 | 96.7% | #7 of 18 | Official leaderboard | |
| DAXBench v2 | 82.2% | #19 of 47 | Official leaderboard | |
| SWE-bench Verified Extended | 59% | #18 of 44 | Official leaderboard | |
| WeirdML v3 | 15.1% | #12 of 27 | Official leaderboard | |
| FrontierCode 2.0 | 25.3% | #15 of 32 | Official leaderboard | |
| MirrorCode | 45.6% | #20 of 39 | Epoch AI run | |
| FrontierCS 2 | 22.9% | #10 of 19 | Official leaderboard | |
| DesignArena Agentic Overall | 1,162 | #41 of 74 | Official leaderboard | |
| DesignArena Blender | 1,158 | #7 of 12 | Official leaderboard | |
| DeepRSI | 955 | #5 of 8 | Official leaderboard | |
| Web Clone Bench | 3.70% | #7 of 10 | Official leaderboard | |
| Android Bench 2.0 | 13.3% | #12 of 17 | Official leaderboard | |
| OpenDesign Arena | 76.2 | #16 of 21 | Official leaderboard | |
| PostTrainBench 1.2 | 24.2% | #17 of 21 | Official leaderboard | |
| FrontierMemory | 51.1% | #6 of 7 | Official leaderboard | |
| MacroscopeBench | 64.2% | #16 of 16 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LisanBench | 27973.3 | #2 of 133 | Official leaderboard | |
| Terminal-Bench 2.1 Extended | 54% | #1 of 44 | Official leaderboard | |
| Mystery Game Puzzles | 51% | #15 of 84 | Epoch AI run | |
| BrowseComp Extended | 47.5% | #8 of 42 | Official leaderboard | |
| DuelLab GameBench 2 | 77.8 | #13 of 57 | Official leaderboard | |
| Terminal-Bench 4.0 | 47% | #10 of 40 | Official leaderboard | |
| Agent Arena | 1.77% | #24 of 56 | Official leaderboard | |
| RuneBench | 4.98 | #34 of 76 | Official leaderboard | |
| EBR-bench | 31.4% | #15 of 31 | Epoch AI run | |
| AutomationBench | 39.1% | #24 of 47 | Lab model card | |
| CUA-bench | 7.17% | #6 of 9 | Official leaderboard | |
| FocusBench (Meta) | 44.9% | #4 of 5 | Official leaderboard | |
| OSWorld 2.1 | 72.6% | #6 of 7 | Lab model card |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| TaxCalcBench v2 | 83.7% | #1 of 32 | Official leaderboard | |
| PRBench | 58.4% | #3 of 48 | Official leaderboard | |
| SnorkelManufacturing | 27.8% | #1 of 16 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 17.1% | #7 of 80 | Official leaderboard | |
| GDPval Extended | 6.60% | #4 of 43 | Official leaderboard | |
| Legal Research Bench | 50.5% | #9 of 93 | Official leaderboard | |
| APEX-Accounting | 67.1% | #5 of 49 | Official leaderboard | |
| Harvey LAB v1.1 (HLAB-Vals-1.1) | 23.8% | #4 of 38 | Official leaderboard | |
| Legora BAR | 105.0 | #2 of 17 | Official leaderboard | |
| Public Benefits Bench | 70.2% | #6 of 50 | Official leaderboard | |
| MedXpertQA MM Extended | 65.7% | #5 of 41 | Official leaderboard | |
| DAYJOB: Law | 36.6% | #2 of 16 | Official leaderboard | |
| Digital Work Index | 26.4 | #2 of 16 | Official leaderboard | |
| SnorkelRevOps | 18% | #2 of 16 | Official leaderboard | |
| Vals Legal Index | 75.3% | #5 of 38 | Official leaderboard | |
| JobBench | 70.5% | #2 of 15 | Lab model card | |
| Tuesday Work Index | 69.7 | #6 of 44 | Official leaderboard | |
| WorkplaceAgents | 17.4% | #2 of 14 | Official leaderboard | |
| APEX-Agents | 66.1% | #10 of 61 | Collected by Epoch AI, source not named | |
| DesignArena Spreadsheet | 1,292 | #6 of 35 | Official leaderboard | |
| DAYJOB: Accounting | 30.8% | #3 of 17 | Official leaderboard | |
| FrontierFinance | 52.6% | #3 of 17 | Official leaderboard | |
| Vending-Bench 2 | $9,432 | #12 of 67 | Official leaderboard | |
| SnorkelLegal | 38.3% | #3 of 16 | Official leaderboard | |
| Tax Agent Bench | 36.3% | #16 of 84 | Official leaderboard | |
| Vals Index v2.1 | 61.1% | #10 of 49 | Official leaderboard | |
| APEX-Agents 1.1 | 77% | #14 of 65 | Official leaderboard | |
| PolicyBench | 90% | #11 of 50 | Official leaderboard | |
| Vals Finance Agent v2 | 56.6% | #18 of 80 | Official leaderboard | |
| Vals Finance Agent v2.1 | 65.5% | #9 of 40 | Official leaderboard | |
| PharmaBench (micro1) | 17.2% | #3 of 13 | Official leaderboard | |
| EnterpriseBench: CoreCraft | 61.5% | #12 of 51 | Official leaderboard | |
| Immigration Bench | 69.6% | #6 of 25 | Official leaderboard | |
| Vals Index v2.2 | 64.9% | #9 of 37 | Official leaderboard | |
| Excel Modeling Benchmark | 67.1% | #19 of 76 | Official leaderboard | |
| micro1 Medical Index | 70.8% | #2 of 8 | Official leaderboard | |
| Vals Finance Index | 61.2% | #11 of 40 | Official leaderboard | |
| LifeSci Diamond | 9.10% | #5 of 18 | Official leaderboard | |
| Excel Modeling Benchmark v1.1 | 72.5% | #12 of 42 | Official leaderboard | |
| RealmPathology | 76.4% | #4 of 13 | Official leaderboard | |
| DesignArena HTML Slides | 1,135 | #9 of 28 | Official leaderboard | |
| EEBench | 62.7% | #10 of 30 | Official leaderboard | |
| DesignArena PDF | 1,200 | #16 of 37 | Official leaderboard | |
| YC-Bench | $1,061,978 | #32 of 59 | Official leaderboard | |
| GDPx | 6.82% | #11 of 20 | Official leaderboard | |
| SnorkelFinance 2.0 | 15% | #9 of 16 | Official leaderboard | |
| SnorkelUnderwrite 2.0 | 23% | #9 of 16 | Official leaderboard | |
| Agents' Last Exam v1.1 | 38.5% | #8 of 14 | Official leaderboard | |
| MissionReport | 62.1% | #6 of 10 | Official leaderboard | |
| HealthBench Professional | 60.5% | #16 of 26 | Lab model card | |
| Harvey LAB-AA v1.1 | 10% | #2 of 3 | Lab model card | |
| DesignArena Python-PPTX Slides | 1,065 | #33 of 36 | Official leaderboard | |
| RWE-bench | 0% | #14 of 15 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MineBench | 1,850 | #14 of 75 | Official leaderboard | |
| Humanity's Sixth Sense | 43.7% | #7 of 34 | Official leaderboard | |
| MMMU-Pro Extended | 43.2% | #10 of 38 | Official leaderboard | |
| VoxelBench | 1,828 | #19 of 56 | Official leaderboard | |
| Furniture Assembly Benchmark | 54.2% | #15 of 39 | Epoch AI run | |
| BVB | 67.4 | #20 of 45 | Official leaderboard | |
| Blueprint-Bench 2 | 31.7% | #19 of 38 | Official leaderboard | |
| CharXiv Extended | 79.6% | #20 of 38 | Official leaderboard | |
| GDP.pdf | 30.8% | #9 of 17 | Lab model card | |
| DataCenterBench Tier 1 | 16.7% | #5 of 9 | Official leaderboard | |
| Dual-Arm Sim | 9.50 | #9 of 11 | Official leaderboard | |
| RoboColiseum | 42.3% | #8 of 9 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CyberBench v1.1 (Vals) | 80.4% | #1 of 44 | Official leaderboard | |
| DistressBench | 85.9% | #3 of 42 | Official leaderboard | |
| BioSecBench-Refusal | 55.3% | #2 of 22 | Official leaderboard | |
| PCT (Polarized Contrastive Pairs) | 33.3% | #5 of 46 | Official leaderboard | |
| nocot-bench (NCRI) | 97.7 | #26 of 185 | Official leaderboard | |
| Biopharma Bench | 61.4% | #3 of 13 | Official leaderboard | |
| HackBench | 61.5 | #2 of 8 | Official leaderboard | |
| OncologyBench | 24% | #8 of 30 | Official leaderboard | |
| CWE-bench 2 | 66.1% | #4 of 14 | Official leaderboard | |
| CheatBench | 12.7% | #3 of 10 | Official leaderboard | |
| VIALS 2 | 31.5% | #6 of 14 | Official leaderboard | |
| WellnessBench | 55% | #7 of 12 | Official leaderboard | |
| FrontierBioML | 26.9 | #7 of 11 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CAIS Risk Index | 32.4 | #6 of 67 | Official leaderboard | |
| CAIS Vision Capabilities Index | 71.2 | #5 of 53 | Official leaderboard | |
| Conceptual Reasoning Index | 67.7 | #15 of 151 | Official leaderboard | |
| LMCA | 56.8% | #15 of 147 | Official leaderboard | |
| ACCoRD | 76.2 | #26 of 152 | Official leaderboard | |
| CAIS Text Capabilities Index | 47.6 | #13 of 68 | Official leaderboard | |
| LiveBench | 70.7% | #10 of 31 | Official leaderboard | |
| Vals Science Index | 39.1% | #12 of 25 | Official leaderboard | |
| Artificial Analysis Intelligence Index v4.3 | 52.0 | #2 of 4 | Lab model card |
Its lab's numbers put it at 163.9, 2.2 points above its ECI
The Self-Reported Index fits one score to the 8 capability results Meta 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 160.4 to 167.5. How we compute it.
- DeepSWE 78.8
- JobBench 70.5
- HealthBench Professional 60.5
- Vals Finance Agent v2 56.6
- SWE-Marathon 53.1
- Terminal-Bench 4.0 52
- AutomationBench 39.1
- GDP.pdf 30.8
Its lab's numbers next to 4 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 |
|---|---|---|---|
| Terminal-Bench 4.0 | 52% Terminal-Bench 4.0-AA | 47% Leaderboard, xhigh, Muse Code | +5.0 pts |
| ProgramBench almost solved | 50% ProgramBench, rollouts passing at least 95% of tests, mini-swe-agent, Muse Wonder (max) (capabilities table image, page 3) | 50.5% Leaderboard, xhigh, mini-SWE-agent | −0.5 pts |
| Vals Finance Agent v2 | 56.6% Vals Finance Agent V2 | 56.6% Leaderboard | Same |
| FocusBench (Meta) | 44.9% FocusBench (chart label) | 44.9% Leaderboard, max | Same |
Meta's models over time
Meta's models in color, against every other model on the Epoch Capabilities Index, by release date.
Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.
Its system card
Every result in the longer documents the lab published with it. 1 of them isn't in the launch post.
Muse Wonder Model Card Oct 12, 2026
Capabilities
- Personal Superintelligence Index (Meta) 54.3
- Harvey LAB-AA v1.1 10
- Vals Finance Agent v2 56.6
- AutomationBench 39.1
- GDP.pdf 30.8
- JobBench 70.5
- OSWorld 2.1 72.6
- MemoryBench (Meta) 85.6
- Terminal-Bench 4.0 52
- DeepSWE 78.8
- SWE-Marathon 53.1
- ProgramBench almost solved 50
- Artificial Analysis Intelligence Index v4.3 52