Introducing Muse Code and Muse Spark 1.2 Aug 5, 2026
Muse Spark 1.2
Muse Spark 1.2 ranks first on 5 of the 112 benchmarks with five or more models scored.
113
Benchmarks with a score
5
Ranked first, out of five or more models
7
In the top three
Meta Aug 2026 Muse Spark 1.2
Closed weights Flagship
155.0
#45 of 256
Epoch Capabilities Index
- DeepSWE 54.9% #17 /28
- ZeroBench 54% #2 /76
- Toolathlon-Verified 75.9% #4 /25
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) | 1,493 | #11 of 335 | Official leaderboard | |
| Longform Creative Writing (EQ-Bench) | 81.5 | #12 of 119 | Official leaderboard | |
| DesignArena ASCII Art | 1,278 | #9 of 77 | Official leaderboard | |
| Judgemark v4 | 79.4 | #6 of 40 | Official leaderboard | |
| Creative Writing v3 | 78.0 | #23 of 120 | Official leaderboard | |
| SimpleQA Verified | 60.3% | #19 of 94 | Epoch AI run | |
| StudyArena Text | 1,259 | #20 of 85 | Official leaderboard | |
| OpenAI MRCR v2 | 59% | #12 of 47 | Official leaderboard | |
| BullshitBench | 40% | #38 of 107 | Official leaderboard | |
| BullshitBench V2 | 50% | #51 of 129 | Official leaderboard | |
| Extended NYT Connections | 79.2% | #45 of 103 | Official leaderboard | |
| Antidote | 1,028 | #17 of 36 | Official leaderboard | |
| Lech Mazur Creative Writing | -0.10 | #27 of 50 | Official leaderboard | |
| Hemingway-bench | 1,013 | #24 of 42 | Official leaderboard | |
| AA-LCR Extended | 61.5% | #29 of 49 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ProofBench | 57% | #8 of 59 | Official leaderboard | |
| DTBench | 94.7% | #27 of 180 | Official leaderboard | |
| SimpleBench | 74.5% | #19 of 106 | Official leaderboard | |
| ProofBench v1.1 | 43% | #34 of 52 | Official leaderboard | |
| BioMysteryBench | 64.8% | #23 of 30 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DesignArena Data Visualization | 1,330 | #8 of 118 | Official leaderboard | |
| DesignArena Website | 1,311 | #14 of 151 | Official leaderboard | |
| DesignArena Code | 1,307 | #19 of 143 | Official leaderboard | |
| DesignArena Video to Website | 1,240 | #3 of 19 | Official leaderboard | |
| IOI (Vals) | 49.5% | #9 of 55 | Official leaderboard | |
| DesignArena UI Component | 1,301 | #23 of 139 | Official leaderboard | |
| DesignArena SVG | 1,274 | #19 of 103 | Official leaderboard | |
| DesignArena Game Dev | 1,294 | #29 of 143 | Official leaderboard | |
| DesignArena 3D Design | 1,286 | #29 of 136 | Official leaderboard | |
| DesignArena Android (Kotlin) | 1,210 | #13 of 46 | Official leaderboard | |
| Vibe Code Bench | 79.1% | #34 of 120 | Official leaderboard | |
| WeirdML | 60.4% | #34 of 112 | Official leaderboard | |
| WebDev Arena | 1,534 | #35 of 105 | arena.ai, via Epoch AI | |
| DesignArena Image to HTML | 1,218 | #18 of 51 | Official leaderboard | |
| GameDevBench | 61% | #6 of 17 | Official leaderboard | |
| DesignArena Image to Frontend | 1,233 | #12 of 31 | Official leaderboard | |
| Code Arena | 1,534 | #50 of 128 | Official leaderboard | |
| DesignArena Full Stack | 1,188 | #20 of 49 | Official leaderboard | |
| DesignArena Frontend Web App | 1,204 | #24 of 53 | Official leaderboard | |
| DesignArena Agentic Overall | 1,178 | #35 of 74 | Official leaderboard | |
| DesignArena React Native | 1,176 | #29 of 56 | Official leaderboard | |
| ProgramBench tests passed | 57.2% | #13 of 25 | Official leaderboard | |
| APEX-SWE | 40.5% | #34 of 65 | Official leaderboard | |
| Code Migration | 29.9% | #43 of 78 | Official leaderboard | |
| ProgramBench almost solved | 4.50% | #14 of 25 | Official leaderboard | |
| DesignArena Agentic Game Dev | 1,127 | #16 of 27 | Official leaderboard | |
| DeepSWE | 54.9% | #17 of 28 | Official leaderboard | |
| ReactBench | 23.6% | #24 of 28 | Official leaderboard | |
| IOI v2 (Vals) | 21.8% | #42 of 45 | Official leaderboard | |
| FrontierSWE | 12% | #24 of 25 | Official leaderboard | |
| ProgramBench | 0% | #24 of 25 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MCP Atlas | 90.3% | #1 of 46 | Lab model card | |
| Toolathlon-Verified | 75.9% | #4 of 25 | Official leaderboard | |
| BrowseComp Extended | 32.5% | #14 of 42 | Official leaderboard | |
| Terminal-Bench 2.1 | 82.9% | #22 of 56 | Lab model card | |
| RuneBench | 5.24 | #30 of 76 | Official leaderboard | |
| Chess Puzzles | 27% | #56 of 93 | Epoch AI run | |
| DuelLab GameBench 2 | 57.3 | #42 of 57 | Official leaderboard | |
| Agent Arena | -8.24% | #49 of 56 | Official leaderboard | |
| OSWorld 2.1 | 15.8% | #4 of 4 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| TaxEval v2 | 80.4% | #1 of 129 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 25.4% | #1 of 80 | Official leaderboard | |
| CorpFin v2 | 70.9% | #5 of 119 | Official leaderboard | |
| Vals Finance Agent v2 | 60.6% | #4 of 80 | Official leaderboard | |
| MedScribe | 90.1% | #5 of 97 | Official leaderboard | |
| APEX-1 | 75.6% | #4 of 71 | Official leaderboard | |
| LegalBench | 85.3% | #23 of 138 | Official leaderboard | |
| Public Benefits Bench | 68.5% | #11 of 50 | Official leaderboard | |
| MedCode | 49.4% | #24 of 96 | Official leaderboard | |
| Tax Agent Bench | 32.8% | #21 of 84 | Official leaderboard | |
| FoodTruck Bench | $41,904 | #16 of 59 | Official leaderboard | |
| Legal Research Bench | 43.8% | #28 of 93 | Official leaderboard | |
| DAYJOB: Healthcare | 8.80% | #12 of 37 | Official leaderboard | |
| Vals Index v2 | 57% | #24 of 65 | Official leaderboard | |
| GDPval-AA v2 | 1,631 | #10 of 27 | Lab model card | |
| MedXpertQA MM Extended | 58.2% | #17 of 41 | Official leaderboard | |
| Tuesday Work Index | 53.3 | #20 of 44 | Official leaderboard | |
| MortgageTax | 65.4% | #40 of 87 | Official leaderboard | |
| DAYJOB: Finance | 6.80% | #21 of 39 | Official leaderboard | |
| DesignArena Python-PPTX Slides | 1,154 | #20 of 36 | Official leaderboard | |
| GDPval Extended | 0.90% | #24 of 43 | Official leaderboard | |
| Excel Modeling Benchmark | 57% | #44 of 76 | Official leaderboard | |
| Immigration Bench | 66.5% | #15 of 25 | Official leaderboard | |
| Vals Index v2.1 | 49.3% | #33 of 49 | Official leaderboard | |
| APEX-Agents | 45.4% | #43 of 61 | Collected by Epoch AI, source not named | |
| DesignArena Spreadsheet | 1,093 | #25 of 35 | Official leaderboard | |
| APEX-Agents 1.1 | 59.8% | #48 of 65 | Official leaderboard | |
| TaxCalcBench v2 | 20% | #25 of 32 | Official leaderboard | |
| DesignArena HTML Slides | 1,030 | #24 of 28 | Official leaderboard | |
| DesignArena PDF | 1,116 | #32 of 37 | Official leaderboard | |
| LifeSci Diamond | 0% | #17 of 18 | Official leaderboard | |
| YC-Bench | $9,070 | #59 of 59 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ZeroBench | 54% | #2 of 76 | Official leaderboard | |
| CharXiv Extended | 87.1% | #1 of 38 | Official leaderboard | |
| Vision Arena | 1,294 | #8 of 136 | Official leaderboard | |
| Chartography | 32.1% | #13 of 43 | Official leaderboard | |
| MineBench | 1,552 | #35 of 75 | Official leaderboard | |
| MMMU-Pro Extended | 30.2% | #26 of 38 | Official leaderboard | |
| BVB | 63.4 | #32 of 45 | Official leaderboard | |
| DesignArena Code to Video | 1,103 | #24 of 29 | Official leaderboard | |
| Humanity's Sixth Sense | 24.5% | #30 of 34 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DistressBench | 88.4% | #1 of 42 | Official leaderboard | |
| nocot-bench (NCRI) | 97.1 | #27 of 185 | Official leaderboard | |
| CyberBench v1.1 (Vals) | 69.5% | #21 of 44 | Official leaderboard | |
| SecIT Bench | 69.3% | #18 of 26 | Official leaderboard | |
| VIALS 2 | 22.6% | #10 of 14 | Official leaderboard | |
| SRE-Bench | 0% | #35 of 35 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ACCoRD | 77.4 | #22 of 152 | Official leaderboard | |
| Vals Multimodal Index | 69.8% | #6 of 33 | Official leaderboard | |
| Conceptual Reasoning Index | 62.7 | #31 of 151 | Official leaderboard | |
| LMCA | 48.4% | #34 of 147 | Official leaderboard | |
| SAGE (Vals) | 47.7% | #30 of 81 | Official leaderboard | |
| Xbench | 26.9 | #11 of 17 | Official leaderboard |
Its lab's number next to a leaderboard
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.
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.