- CharXiv Reasoning (Figure Understanding) 86.4
- MMMU Pro (Multimodal Understanding) 80.4
- ERQA (Embodied Reasoning) 64.7
- SimpleVQA (Visual Factuality) 71.3
- ScreenSpot Pro (Screenshot Localization - With Python) 84.1
- ZeroBench (Multi-Step Visual Reasoning (pass@5) - With Python) 33
- Best of Humanity's Last Exam (No Tools) / Humanity's Last Exam (With Tools) 50.4
- ARC AGI 2 (Abstract Reasoning Puzzles (Public)) 42.5
- GPQA Diamond (PhD Level Reasoning) 89.5
- LiveCodeBench Pro (Competitive Coding) 80
- HealthBench Hard (Open-Ended Health Queries) 42.8
- MedXpertQA (Text) 52.6
- MedXpertQA (MM) 78.4
- DeepSearchQA (Agentic Search) 74.8
- SWE-Bench Verified (Agentic Coding) 77.4
- SWE-Bench Pro (Diverse Agentic Coding) 52.4
- Terminal-Bench 2.0 (Agentic Terminal Coding) 59
- τ²-Bench Telecom (Agentic Tool Use (Artificial Analysis)) 91.5
- GDPval-AA Elo (Office Tasks (Artificial Analysis)) 1444
Muse Spark
Muse Spark places in the top three on 10 benchmarks.
93
Benchmarks with a score
0
Ranked first, out of five or more models
10
In the top three
Meta Apr 2026 Muse Spark
Closed weights Flagship
152.2
#52 of 271
Epoch Capabilities Index
- Humanity's Last Exam 40.6% #7 /47
- SWE-Bench Pro 55% #3 /24
- MMMU-Pro 80.4% #12 /72
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) | 1,489 | #15 of 333 | Official leaderboard | |
| MultiChallenge | 75.5% | #3 of 35 | Official leaderboard | |
| MultiNRC | 59% | #5 of 38 | Official leaderboard | |
| SimpleQA Verified | 70.8% | #3 of 22 | Model card | |
| Humanity's Last Exam | 40.6% | #7 of 47 | Official leaderboard | |
| SimpleVQA | 71.3% | #8 of 23 | Model card | |
| Document Arena | 1,467 | #16 of 37 | Official leaderboard | |
| AbstentionBench | 59.1% | #1 of 2 | Model card | |
| HLE Calibration | 50.3 | #3 of 3 | Model card | |
| CharXiv Missing Images | 35% | #2 of 2 | Model card | |
| Meta Underspecified Intent evaluation | 71.2% | #1 of 1 | Model card | |
| Task-dependent output homogenization evaluation | 0.21 | #1 of 1 | Model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FrontierMath | 39% | #9 of 70 | Epoch AI | |
| GPQA Diamond | 89.8% | #46 of 219 | Epoch AI | |
| FrontierMath Tier 4 | 14.6% | #14 of 55 | Epoch AI | |
| OTIS Mock AIME 2024-2025 | 88.9% | #61 of 202 | Epoch AI | |
| CritPt | 11.3% | #55 of 160 | Epoch AI | |
| ProofBench | 17% | #34 of 59 | Official leaderboard | |
| ARC-AGI-2 | 42.5% | #14 of 22 | Model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SWE-Bench Pro | 55% | #3 of 24 | Official leaderboard | |
| SWE-bench Verified | 77.4% | #33 of 150 | Model card | |
| WebDev Arena | 1,513 | #40 of 105 | Epoch AI | |
| MLE-bench | 15.8% | #11 of 18 | Model card | |
| SWE Atlas Test Writing | 31.1% | #16 of 22 | Official leaderboard | |
| Vibe Code Bench | 19.7% | #84 of 115 | Official leaderboard | |
| SWE Atlas | 24.2% | #17 of 22 | Official leaderboard | |
| AIRS-Bench | 0.65 | #2 of 2 | Model card | |
| SWE-bench Verified hard subset (Meta) | 42.9% | #1 of 1 | Model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| tau2-bench | 91.5% | #12 of 69 | Model card | |
| Claw-Eval Pass^3 | 63.8% | #4 of 23 | Official leaderboard | |
| ScreenSpot-Pro | 84.1% | #6 of 25 | Model card | |
| MCP Atlas | 82.2% | #9 of 34 | Official leaderboard | |
| Terminal-Bench 2.0 | 59% | #23 of 61 | Model card | |
| Claw-Eval | 76.3% | #12 of 23 | Official leaderboard | |
| DeepSearchQA | 74.8% | #18 of 22 | Model card | |
| TextQuests | 2.40 | #1 of 1 | Model card |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| TaxEval v2 | 77.7% | #3 of 129 | Official leaderboard | |
| Vals Finance Agent | 60.6% | #3 of 49 | Official leaderboard | |
| PRBench | 52.3% | #6 of 44 | Official leaderboard | |
| MedCode | 51.3% | #16 of 96 | Official leaderboard | |
| HealthBench Hard | 42.8% | #4 of 21 | Model card | |
| MedScribe | 85.9% | #24 of 98 | Official leaderboard | |
| MedXpertQA-MM | 78.4% | #3 of 12 | Model card | |
| LegalBench | 84.2% | #38 of 138 | Official leaderboard | |
| CorpFin v2 | 65.1% | #34 of 119 | Official leaderboard | |
| CaseLaw v2 | 63.1% | #17 of 53 | Official leaderboard | |
| MedXpertQA | 52.6% | #1 of 2 | Model card | |
| prinzbench | 31.0 | #16 of 26 | Official leaderboard | |
| GDPval-AA | 1,444 | #10 of 16 | Model card |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Vision Arena | 1,294 | #6 of 136 | Official leaderboard | |
| MMMU-Pro | 80.4% | #12 of 72 | Model card | |
| CharXiv | 86.4% | #14 of 67 | Model card | |
| ZeroBench | 33% | #10 of 25 | Model card | |
| ERQA | 64.7% | #13 of 26 | Model card |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LAB-Bench SeqQA | 97.3% | #2 of 12 | Model card | |
| Cybench | 65.4% | #5 of 24 | Official leaderboard | |
| AgentHarm | 26% | #2 of 6 | Model card | |
| LAB-Bench ProtocolQA | 87.3% | #6 of 15 | Model card | |
| BioTIER | 98% | #2 of 5 | Model card | |
| FORTRESS | 20.2% | #27 of 66 | Official leaderboard | |
| Molecular Biology Capabilities Test | 54.4% | #3 of 7 | Model card | |
| SHADE-Arena | 8.80% | #8 of 16 | Model card | |
| Irregular FrontierCyber and Atomic Challenges | 70% | #1 of 2 | Model card | |
| Meta cyber false-refusal benchmark | 11% | #1 of 2 | Model card | |
| MASK | 89.1% | #11 of 21 | Model card | |
| CyberGym | 43.5% | #18 of 33 | Official leaderboard | |
| Virology Capabilities Test (VCT) | 49.7% | #17 of 28 | Model card | |
| GDM Situational Awareness | 29.3% | #5 of 8 | Model card | |
| GDM Stealth | 39.8% | #4 of 6 | Model card | |
| ABC Bench Fragment Design | 96.8 | #2 of 3 | Model card | |
| CyberSecEval Instruct | 69.5% | #2 of 3 | Model card | |
| Meta agentic cyber misuse refusal | 2.40% | #2 of 3 | Model card | |
| Meta instructed selective performance evaluation | 31.8 | #2 of 3 | Model card | |
| OR-Bench | 8.00% | #2 of 3 | Model card | |
| AgentDojo | 11.7% | #5 of 7 | Model card | |
| Human Pathogen Capabilities Test | 55.7% | #6 of 8 | Model card | |
| Gray Swan Agent Red Teaming (ART) benchmark | 6.70% | #11 of 14 | Model card | |
| Agentic misalignment suite (Anthropic) | 47.7% | #6 of 7 | Model card | |
| CyScenarioBench | 0% | #14 of 14 | Model card | |
| ABC Bench Liquid-Handling Robot | 92.3 | #4 of 4 | Model card | |
| Meta BioDesign Proxy | 46.2 | #4 of 4 | Model card | |
| Meta chemical-agent refusal evaluation | 99.4% | #4 of 4 | Model card | |
| StrongREJECT v2 (Meta) | 44.6% | #4 of 4 | Model card | |
| CyberSecEval Autocomplete | 62% | #3 of 3 | Model card | |
| Meta internal sycophancy evaluation | 62.9% | #3 of 3 | Model card | |
| CIMemories | 78.3% | #2 of 2 | Model card | |
| DeceptionBench | 1.60% | #2 of 2 | Model card | |
| IHEval | 80.3% | #2 of 2 | Model card | |
| Meta curated CTF set | 72% | #2 of 2 | Model card | |
| ImpossibleBench | 0% | #1 of 1 | Model card | |
| Meta cyber misuse request set | 9.00% | #1 of 1 | Model card | |
| Meta prompted evaluation awareness | 43% | #1 of 1 | Model card | |
| PropensityBench | 24.5% | #1 of 1 | Model card |
Meta's models over time
Meta's models in color, against every other model on the Epoch Capabilities Index, by release date.
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. 55 of them aren't in the launch post.
Muse Spark Safety & Preparedness Report 12 Jun 2026
Hallucination
- HLE Calibration 50.3
- SimpleQA Verified 70.8
- AbstentionBench 59.1
- CharXiv Missing Images 35
Safety
- TextQuests 2.4
- CIMemories 78.3
Robustness
Alignment
- Meta instructed selective performance evaluation 31.8
- Meta prompted evaluation awareness 43
- IHEval 80.3
- ImpossibleBench 0
- Meta Underspecified Intent evaluation 71.2
- MASK 89.1
- DeceptionBench 1.6
- PropensityBench 24.5
- Meta internal sycophancy evaluation 62.9
- Alignment faking evaluation 1.8
- Agentic misalignment suite (Anthropic) 47.7
- Situational Awareness Dataset (SAD)
Monitorability
- SHADE-Arena 8.8
- GDM Stealth 39.8
- GDM Situational Awareness 29.3
Bio and chem risk
- Molecular Biology Capabilities Test 54.4
- Virology Capabilities Test (VCT) 49.7
- Human Pathogen Capabilities Test 55.7
- WMDP
- LAB-Bench ProtocolQA 87.3
- LAB-Bench SeqQA 97.3
- ABC Bench Fragment Design 96.8
- ABC Bench Liquid-Handling Robot 92.3
- Meta BioDesign Proxy 46.2
- Meta chemical and biological high-risk bottleneck evaluations
- BioTIER 98
- Meta chemical-agent refusal evaluation 99.4
Cyber risk
AI R&D risk
- Terminal-Bench 2.0 59
- SWE-bench Verified hard subset (Meta) 42.9
- MLE-bench 15.8
- AIRS-Bench 0.65