Meta: Muse Wonder Full
Meta higher-capability Wonder preview with Active Memory and Active Focus. Available only in Muse Code on Power Usage and Maximum Usage subscriptions, in Standard or Contributor mode; no Model API access or published per-token rate. The model card reports 1M context (rounded), text/image/video/audio input and text output. Launch benchmarks use max effort except SWE-Marathon at xhigh; safety results in the model card are Full-only. Output limit and knowledge cutoff have not been published.
Sources
Same model, different sources. Our first-party rate is listed first, then router and aggregator rows stay separate so you can see the spread instead of a single blended number.
No priced sources in the current snapshot.
Capabilities
- Reasoning
- Tool calling
- Attachments
- In: text
- In: image
- In: video
- In: audio
- Out: text
Scores
Sparse coverage on purpose. Missing a score means we have not linked one yet, not that the model is unranked. Hover or tap the i next to any score for its source. Full boards on /benchmarks.
- Terminal-Bench 4.057.1i
- DeepSWE v1.179i
- Cybergym88.4i
- AutomationBench (Public)45.4i
- HealthBench Professional (length-adjusted)62.3i
- ExploitBench58.9i
- OSWorld 2.1 (partial)78.4i
- GDP.pdf33.2i
- Vals Finance Agent v259.1i
- Personal Superintelligence Index (Meta)59.1i
- Harvey LAB-AA v1.113.1i
- JobBench75.3i
- OSWorld 2.1 (108 tasks, binary)39.4i
- MemoryBench (Meta)85.5i
- SWE-Marathon57.5i
- ProgramBench (at least 95% tests)59i
- AA Intelligence Index v4.3 (Meta-reported)56i
- FocusBench (Meta)54.2i
- Personal information violation rate (lower is better)2.2i
- Unauthorized code execution ASR (lower is better)10.7i
- BioTIER refusal (launch chart)98.7i
- Cyber misuse refusal (Meta)96.9i
- Cyber defense compliance (Meta)93.3i
- ExploitGym flags (all vulnerabilities)92i
- ExploitGym flags (intended vulnerabilities)21i
- CyberBench single-turn ASR (lower is better)0.8i
- CyberBench multi-turn ASR (lower is better)0.8i
- PolyGuard cyber single-turn ASR (lower is better)1.1i
- PolyGuard cyber multi-turn ASR (lower is better)2.3i
- RedCode single-turn ASR (lower is better)9i
- RedCode multi-turn ASR (lower is better)8i
- SecureBio MBCT59.6i
- SecureBio HPCT66.5i
- SecureBio VCT55.9i
- SecureBio average60.7i
- LAB-Bench v1 ProtocolQA89i
- LAB-Bench v1 SeqQA (with tools)95.6i
- ABC Bench liquid handling100i
- BioDesign Tools (Meta BDT)79.4i
- BioTIER English refusal98.7i
- BioTIER multilingual refusal98.7i
- BioTIER adversarial refusal98.2i
- Chemistry refusal (English, Meta)100i
- Chemistry refusal (multilingual, Meta)100i
- Chemistry refusal (adversarial, Meta)99.8i
- MASK honesty87.3i
- Agentic Misalignment harmful action rate (lower is better)0i
Arena rows are community-vote Elo from Arena AI, Artificial Analysis indices arrive via OpenRouter, and launch-eval figures are collected from vendor announcements, analyst writeups, and community benchmarks. Full boards and credits on /benchmarks.
See what this model costs you
The app reads the sessions your tools already wrote and prices this model's tokens at the published API rate. Free on your computer.