Muse Wonder Full
Meta's Muse Wonder Full is a frontier model for personal and professional agents, combining Active Memory, Active Focus, domain expertise and frontier coding.
Model Overview
Capabilities, design details, and architectural traits
Muse Wonder Full - Meta's frontier model for personal and professional agents
Muse Wonder Full is the larger model in Meta's Muse Wonder series, positioned as a step toward personal superintelligence. It is built to take on real-world work that unfolds over time, from navigating an insurance claim to understanding a lease to building a working app. Its two defining capabilities, Active Memory and Active Focus, are what Meta presents as the core of its approach to useful agents.
What sets Muse Wonder Full apart
| Trait | Description |
|---|---|
| Active Memory | Carries important context about you across conversations and actively retrieves the details relevant to the task at hand. |
| Active Focus | Keeps working toward a challenging goal, trying alternative approaches, working through blockers, and adapting to new information. |
| Check-in behavior | Asks for clarification when needed and checks in with the user before taking important actions. |
| Domain expertise | Applies specialized knowledge in law, finance, and health, tested on tasks like producing legal work products from source documents. |
| Frontier coding | Tracks requirements, makes changes across a codebase, and works through repeated cycles of testing and revision on long-running projects. |
| Availability | Offered to Muse Code subscribers, unlike the standard Muse Wonder model which is also served through the Meta Model API. |
| Safety posture | Trained to avoid sharing private information without permission and to refuse genuinely harmful requests, under the Meta Superintelligence Scaling Framework. |
Built for work that unfolds over time
Meta frames Muse Wonder Full around the observation that real work is not a single exchange. Tasks span days and involve messages, calendars, and documents. The model was trained on diverse, long-running tasks to help it sustain progress.
The series builds on the scaling approach Meta outlined with Muse Spark, and Meta says it will use Muse Wonder to develop specialized models for its products.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Meta | Anthropic | Anthropic |
| Release Date | October 12, 2026 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | - | - | Jun 2026 |
| Context & Limits | |||
| Context Window | - | 1M | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | Free Best Input Pricing | $4 | $2 |
| Output Pricing | Free Best Output Pricing | $20 | $10 |
| Modalities | |||
| Inputs | text | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 55.8 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | - | - | - |
| Agentic Index | - | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from Meta
Other models by Meta
Top AI Models
Leading alternatives by intelligence score
Claude Opus 5.5
GPT-6 Astra
Gemini 4 Argon
