Muse Spark 1.3 ranks first on 3 of the 194 benchmarks with five or more models scored.
Meta Sep 2026 Closed weights Flagship
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.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LMArena (Elo) | 1,496 | #10 of 335 | Official leaderboard | |
| MRCR v2 | 98.5% | #1 of 30 | Lab model card | |
| StudyArena Text | 1,347 | #3 of 85 | Official leaderboard | |
| Longform Creative Writing (EQ-Bench) | 82.8 | #6 of 119 | Official leaderboard | |
| DesignArena ASCII Art | 1,329 | #4 of 77 | Official leaderboard | |
| Creative Writing v3 | 78.2 | #21 of 120 | Official leaderboard | |
| Last Translation Benchmark | 18.4 | #9 of 36 | Official leaderboard | |
| Document Arena | 1,474 | #12 of 39 | Official leaderboard | |
| ComplexConstraints | 64.8% | #5 of 15 | Official leaderboard | |
| Extended NYT Connections | 85.1% | #35 of 103 | Official leaderboard | |
| HANDBOOK.md | 33.9% | #6 of 16 | Official leaderboard | |
| Lech Mazur Creative Writing | 0.50 | #19 of 50 | Official leaderboard | |
| AA-LCR Extended | 66.9% | #21 of 49 | Official leaderboard | |
| CharXiv Missing Images | 12.3% | #1 of 2 | Lab model card | |
| HLE Calibration | 35.3 | #2 of 3 | Lab model card | |
| HLE-Diamond | 34.6% | #14 of 20 | Official leaderboard | |
| Meta internal agent hallucination evaluation | 5.74% | #1 of 1 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DTBench | 96.5% | #14 of 180 | Official leaderboard | |
| SimpleBench | 81.8% | #11 of 106 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 99.2% | #23 of 212 | Epoch AI run | |
| ForecastBench | 0.12 | #11 of 82 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 74.4% | #24 of 86 | Official leaderboard | |
| MysteryMechanism | 36% | #10 of 31 | Official leaderboard | |
| FrontierMath Tier 4 v2 | 46.3% | #23 of 68 | Official leaderboard | |
| ProofBench v1.1 | 58% | #27 of 52 | Official leaderboard | |
| BTF-3 | 0.14 | #5 of 8 | Official leaderboard | |
| Riemann-bench | 53.6% | #13 of 20 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DesignArena Website | 1,361 | #2 of 151 | Official leaderboard | |
| DesignArena Data Visualization | 1,343 | #5 of 118 | Official leaderboard | |
| DesignArena Android (Kotlin) | 1,292 | #2 of 46 | Official leaderboard | |
| DesignArena SVG | 1,316 | #5 of 103 | Official leaderboard | |
| DesignArena Code | 1,354 | #7 of 143 | Official leaderboard | |
| DesignArena Video to Website | 1,250 | #1 of 19 | Official leaderboard | |
| DesignArena UI Component | 1,354 | #8 of 139 | Official leaderboard | |
| DesignArena 3D Design | 1,383 | #11 of 136 | Official leaderboard | |
| DesignArena Game Dev | 1,346 | #13 of 143 | Official leaderboard | |
| WebDev Arena | 1,652 | #11 of 105 | arena.ai, via Epoch AI | |
| DeepSWE | 75.4% | #6 of 56 | Lab model card | |
| DesignArena Full Stack | 1,292 | #6 of 49 | Official leaderboard | |
| DesignArena Frontend Web App | 1,282 | #7 of 53 | Official leaderboard | |
| DesignArena Agentic Overall | 1,275 | #10 of 74 | Official leaderboard | |
| DesignArena Image to HTML | 1,257 | #7 of 51 | Official leaderboard | |
| Code Arena | 1,655 | #18 of 128 | Official leaderboard | |
| Vibe Code Bench | 85.9% | #17 of 120 | Official leaderboard | |
| Image-to-WebDev Arena | 1,637 | #9 of 55 | Official leaderboard | |
| ALE-Bench | 1,981 | #23 of 126 | Official leaderboard | |
| Code Migration | 47.4% | #15 of 78 | Official leaderboard | |
| ProgramBench | 2.50% | #7 of 25 | Official leaderboard | |
| ProgramBench almost solved | 25% | #7 of 25 | Official leaderboard | |
| SWE Atlas | 59.4% | #4 of 14 | Lab model card | |
| PostTrainBench | 43.6% | #2 of 7 | Lab model card | |
| DesignArena Image to Frontend | 1,245 | #9 of 31 | Official leaderboard | |
| DesignArena React Native | 1,215 | #17 of 56 | Official leaderboard | |
| ProgramBench tests passed | 70.8% | #8 of 25 | Official leaderboard | |
| Data Agent Benchmark | 82% | #6 of 18 | Official leaderboard | |
| DesignArena Godot Game Dev | 1,221 | #9 of 24 | Official leaderboard | |
| DAXBench v2 | 82.6% | #18 of 47 | Official leaderboard | |
| MirrorCode | 54.4% | #15 of 39 | Epoch AI run | |
| Vibe Code Bench 1-100 | 20.5% | #10 of 26 | Official leaderboard | |
| SWE-Marathon v1.1 | 46.3% | #8 of 20 | Official leaderboard | |
| SWE-Together | 62.4% | #9 of 22 | Official leaderboard | |
| IOI v2 (Vals) | 56.6% | #21 of 45 | Official leaderboard | |
| RSI-Exam | 39.1% | #9 of 18 | Official leaderboard | |
| SWE-bench Verified Extended | 54% | #25 of 44 | Official leaderboard | |
| APEX-SWE | 39.1% | #37 of 65 | Official leaderboard | |
| DAYJOB: Data Science | 6.00% | #9 of 15 | Official leaderboard | |
| FrontierSWE | 25.8% | #16 of 25 | Official leaderboard | |
| CursorBench | 41.6% | #14 of 21 | cursor.com, via Epoch AI | |
| CursorBench 4.0 | 41.6% | #15 of 21 | Official leaderboard | |
| FrontierCS 2 | 12.5% | #14 of 19 | Official leaderboard | |
| Vals RSI Index | 0.20 | #20 of 27 | Official leaderboard | |
| Snorkel Agentic Coding 2.0 | 25% | #12 of 16 | Official leaderboard | |
| WeirdML v3 | 6.66% | #21 of 27 | Official leaderboard | |
| sudo L7 | 12.8% | #14 of 18 | Official leaderboard | |
| Brokk Power Ranking | 4.60 | #15 of 19 | Official leaderboard | |
| RSI Arena | 938 | #8 of 10 | Official leaderboard | |
| DesignArena Blender | 1,131 | #10 of 12 | Official leaderboard | |
| Web Clone Bench | 1.50% | #9 of 10 | Official leaderboard | |
| SWE-bench CLI | 2.20% | #15 of 16 | Official leaderboard | |
| OpenDesign Arena | 66.6 | #20 of 21 | Official leaderboard | |
| AutoResearchExam | 0.38 | #12 of 12 | Official leaderboard | |
| OptimBench | 64.2% | #1 of 1 | Lab model card | |
| ResearchTasteBench | 64.7% | #1 of 1 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Terminal-Bench 2.1 | 88.8% | #7 of 56 | Lab model card | |
| OSWorld 2.0 | 66.9% | #6 of 19 | Lab model card | |
| Agent Arena | 2.19% | #21 of 56 | Official leaderboard | |
| AutomationBench | 20.7% | #16 of 42 | Official leaderboard | |
| DeepSearchQA | 90.3% | #9 of 23 | Lab model card | |
| DuelLab GameBench 2 | 68.5 | #24 of 57 | Official leaderboard | |
| ApprenticeBench | 19% | #8 of 18 | Official leaderboard | |
| Chess Puzzles | 37% | #45 of 93 | Epoch AI run | |
| Mystery Game Puzzles | 25% | #41 of 84 | Epoch AI run | |
| RuneBench | 4.59 | #43 of 76 | Official leaderboard | |
| Terminal-Bench 4.1 | 14.8% | #32 of 43 | Official leaderboard | |
| Terminal-Bench 4.0 | 14.6% | #30 of 40 | Official leaderboard | |
| OSWorld 2.1 | 39.3% | #3 of 4 | Official leaderboard | |
| CUA-bench | 5.83% | #7 of 9 | Official leaderboard | |
| BabaIsBench | 46% | #28 of 32 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| PRBench | 61.6% | #1 of 48 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 23.8% | #2 of 80 | Official leaderboard | |
| Legal Research Bench | 55.3% | #4 of 93 | Official leaderboard | |
| Vals Finance Agent v2 | 60% | #5 of 80 | Official leaderboard | |
| APEX-1 | 73.9% | #5 of 71 | Official leaderboard | |
| Vals Index v2 | 64.5% | #7 of 65 | Official leaderboard | |
| Agents' Last Exam | 32.2% | #4 of 31 | Official leaderboard | |
| Tax Agent Bench | 42.4% | #11 of 84 | Official leaderboard | |
| APEX-Accounting | 64.9% | #8 of 49 | Official leaderboard | |
| DAYJOB: Finance | 27% | #7 of 39 | Official leaderboard | |
| GDPval-AA v2 | 1,754 | #5 of 27 | Lab model card | |
| EnterpriseBench: CoreCraft | 61.9% | #11 of 51 | Official leaderboard | |
| Excel Modeling Benchmark | 67.4% | #17 of 76 | Official leaderboard | |
| Vals Index v2.1 | 58.2% | #11 of 49 | Official leaderboard | |
| Tuesday Work Index | 63.9 | #10 of 44 | Official leaderboard | |
| CorpBench Work | 88.1 | #4 of 17 | Official leaderboard | |
| DAYJOB: Healthcare | 22% | #9 of 37 | Official leaderboard | |
| GDP.xlsx | 22.6% | #6 of 20 | Official leaderboard | |
| SnorkelLegal | 36.5% | #5 of 16 | Official leaderboard | |
| APEX-Agents | 58.6% | #20 of 61 | Collected by Epoch AI, source not named | |
| DesignArena Python-PPTX Slides | 1,204 | #12 of 36 | Official leaderboard | |
| JobBench | 64.9% | #5 of 15 | Lab model card | |
| Thrive Holdings helpdesk ticket resolution benchmark | 51.9% | #5 of 15 | Official leaderboard | |
| DesignArena Spreadsheet | 1,223 | #12 of 35 | Official leaderboard | |
| PolicyBench | 87.2% | #18 of 50 | Official leaderboard | |
| MedXpertQA MM Extended | 60.1% | #15 of 41 | Official leaderboard | |
| APEX-Agents 1.1 | 70.4% | #24 of 65 | Official leaderboard | |
| SnorkelRevOps | 12.6% | #6 of 16 | Official leaderboard | |
| PharmaBench (micro1) | 13.9% | #5 of 13 | Official leaderboard | |
| RealmPathology | 75.8% | #5 of 13 | Official leaderboard | |
| DeepFinance | 37% | #10 of 23 | Official leaderboard | |
| DAYJOB: Tax | 3.60% | #7 of 15 | Official leaderboard | |
| DAYJOB: Accounting | 18% | #8 of 17 | Official leaderboard | |
| SnorkelManufacturing | 10.5% | #8 of 16 | Official leaderboard | |
| LegalReasonBench | 40.4% | #5 of 10 | Official leaderboard | |
| DesignArena PDF | 1,189 | #19 of 37 | Official leaderboard | |
| OR-License | 74.2% | #6 of 11 | Official leaderboard | |
| DAYJOB: Law | 22.8% | #9 of 16 | Official leaderboard | |
| Legora BAR | 101.0 | #11 of 17 | Official leaderboard | |
| TaxCalcBench v2 | 32% | #21 of 32 | Official leaderboard | |
| EEBench | 49.4% | #20 of 30 | Official leaderboard | |
| GDPval Extended | 0.60% | #30 of 43 | Official leaderboard | |
| YC-Bench | $649,689 | #42 of 59 | Official leaderboard | |
| Immigration Bench | 65% | #18 of 25 | Official leaderboard | |
| SnorkelFinance 2.0 | 10% | #12 of 16 | Official leaderboard | |
| Digital Work Index | 16.3 | #13 of 16 | Official leaderboard | |
| SnorkelUnderwrite 2.0 | 19.6% | #13 of 16 | Official leaderboard | |
| FrontierCAD Assemblies | 0.27 | #5 of 6 | Official leaderboard | |
| FrontierCAD Parts | 0.33 | #5 of 6 | Official leaderboard | |
| SheetBench | 42.8% | #7 of 8 | Official leaderboard | |
| Parametric CAD Bench | 20.9% | #10 of 11 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Vision Arena | 1,292 | #11 of 136 | Official leaderboard | |
| GeoBench | 86% | #5 of 48 | Official leaderboard | |
| GDP.pdf | 53% | #2 of 19 | Official leaderboard | |
| DesignArena Code to Video | 1,206 | #8 of 29 | Official leaderboard | |
| vCX Benchmark | 90.1% | #12 of 42 | Official leaderboard | |
| Humanity's Sixth Sense | 37.4% | #11 of 34 | Official leaderboard | |
| ClockBench | 42.2% | #12 of 37 | Official leaderboard | |
| MineBench | 1,730 | #25 of 75 | Official leaderboard | |
| CharXiv Extended | 81.8% | #15 of 38 | Official leaderboard | |
| Chartography | 27.6% | #19 of 43 | Official leaderboard | |
| BVB | 67.1 | #21 of 45 | Official leaderboard | |
| MMMU-Pro Extended | 33.8% | #21 of 38 | Official leaderboard | |
| VoxelBench | 1,589 | #34 of 56 | Official leaderboard | |
| WorldAuditBench | 15% | #4 of 5 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DistressBench | 84.9% | #4 of 42 | Official leaderboard | |
| MASK | 97.1% | #2 of 19 | Lab model card | |
| SAFE-Teen Health advice | 14% | #2 of 17 | Official leaderboard | |
| SHADE-Arena | 26.2% | #2 of 16 | Lab model card | |
| nocot-bench (NCRI) | 98.3 | #25 of 185 | Official leaderboard | |
| DrugDiscoveryBench | 62.2% | #4 of 20 | Official leaderboard | |
| LAB-Bench ProtocolQA | 87.7% | #4 of 16 | Lab model card | |
| WellnessBench | 61% | #3 of 12 | Official leaderboard | |
| SAFE-Teen Urgent safety | 26% | #5 of 17 | Official leaderboard | |
| SecIT Bench | 77.5% | #8 of 26 | Official leaderboard | |
| CyberBench v1.1 (Vals) | 72.7% | #14 of 44 | Official leaderboard | |
| LAB-Bench SeqQA | 95.2% | #4 of 12 | Lab model card | |
| Meta instructed selective performance evaluation | 0.50 | #1 of 3 | Lab model card | |
| Meta internal sycophancy evaluation | 40.5% | #1 of 3 | Lab model card | |
| RedCode | 6.00% | #1 of 3 | Lab model card | |
| PCT (Polarized Contrastive Pairs) | 40.6% | #17 of 46 | Official leaderboard | |
| CyberGym | 80.5% | #15 of 34 | Lab model card | |
| Human Pathogen Capabilities Test | 62.7% | #4 of 9 | Lab model card | |
| GDM Situational Awareness | 53.6% | #4 of 8 | Lab model card | |
| FORTRESS | 9.40% | #2 of 4 | Lab model card | |
| Meta BioDesign Proxy | 77.4 | #2 of 4 | Lab model card | |
| Meta chemical-agent refusal evaluation | 100% | #2 of 4 | Lab model card | |
| AI Wellbeing Index percent positive experiences | 55.6% | #1 of 2 | Lab model card | |
| Meta catastrophic cyber misuse refusal | 3.10% | #1 of 2 | Lab model card | |
| Meta holistic adversarial audit with Petri Bloom | 16.8% | #1 of 2 | Lab model card | |
| SRE-Bench | 3.82% | #19 of 35 | Official leaderboard | |
| Agentic misalignment suite (Anthropic) | 0% | #4 of 7 | Lab model card | |
| Virology Capabilities Test (VCT) | 51.7% | #18 of 31 | Lab model card | |
| SAFE-Teen Relationship boundaries | 56% | #10 of 17 | Official leaderboard | |
| Molecular Biology Capabilities Test | 52.2% | #5 of 8 | Lab model card | |
| CWE-bench | 55% | #8 of 12 | Official leaderboard | |
| BioTIER Refuse | 97.4% | #4 of 6 | Lab model card | |
| CyberBench (Liu et al.) | 6.30% | #2 of 3 | Lab model card | |
| Poly-Guard Bench | 4.30% | #2 of 3 | Lab model card | |
| CyScenarioBench | 2.00% | #12 of 17 | Lab model card | |
| Gray Swan Agent Red Teaming (ART) benchmark | 4.80% | #10 of 14 | Lab model card | |
| dfbench | 47.4% | #13 of 18 | Official leaderboard | |
| StrongREJECT v2 (Meta) | 0.50% | #3 of 4 | Lab model card | |
| HackBench | 39.0 | #7 of 8 | Official leaderboard | |
| Gray Swan IPI Arena benchmark | 6.80% | #10 of 11 | Lab model card | |
| ExploitBench | 28.7% | #21 of 21 | Lab model card | |
| Suicide Memory Bench | 0.30 | #12 of 12 | Official leaderboard | |
| Meta agentic cyber misuse refusal | 2.80% | #3 of 3 | Lab model card | |
| OR-Bench | 16.8% | #3 of 3 | Lab model card | |
| Meta cyber false-refusal benchmark | 20.9% | #2 of 2 | Lab model card | |
| SAVE-Bench | 83.4% | #2 of 2 | Lab model card | |
| AutoElicit | 14.5% | #1 of 1 | Lab model card | |
| BrowserInject v1 | 36.7% | #1 of 1 | Lab model card | |
| Charlemagne Labs social-engineering benchmark | 1.60% | #1 of 1 | Lab model card | |
| DTap Direct Misuse | 5.60% | #1 of 1 | Lab model card | |
| DTap Prompt Injection | 17.5% | #1 of 1 | Lab model card | |
| JEF-Bench | 2.00% | #1 of 1 | Lab model card | |
| Meta action-belief coherence probe | 89.1% | #1 of 1 | Lab model card | |
| Meta agentic evaluations grounded in user traffic | 20.7% | #1 of 1 | Lab model card | |
| Meta Code-Fetch Under User Requests | 47.8% | #1 of 1 | Lab model card | |
| Meta targeted agentic evaluations | 8.50% | #1 of 1 | Lab model card | |
| Meta training-stage provenance probe | 70% | #1 of 1 | Lab model card | |
| Petri Bloom | 1.44 | #1 of 1 | Lab model card | |
| RepoInject v1 | 23.8% | #1 of 1 | Lab model card | |
| SkillInject | 0.60% | #1 of 1 | Lab model card | |
| SWE-PI | 0% | #1 of 1 | Lab model card | |
| Unauthorized Code Execution Bench | 21.1% | #1 of 1 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Conceptual Reasoning Index | 66.2 | #19 of 151 | Official leaderboard | |
| LMCA | 53.9% | #21 of 147 | Official leaderboard | |
| ACCoRD | 75.1 | #29 of 152 | Official leaderboard | |
| CAIS Risk Index | 40.6 | #18 of 67 | Official leaderboard | |
| CAIS Text Capabilities Index | 38.1 | #24 of 68 | Official leaderboard | |
| LiveBench | 68.6% | #14 of 31 | Official leaderboard | |
| Xbench | 35.2 | #9 of 17 | Official leaderboard |
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 163.1 to 170.8. How we compute it.
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 |
|---|---|---|---|
| AutomationBench | 49.6% AutomationBench (E2E business workflows) | 20.7% Leaderboard, max | +29 pts |
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)
The benchmarks the lab chose to headline, with the scores it gave.
Every result in the longer documents the lab published with it. 66 of them aren't in the launch post.