What Meta Launched
On October 12, 2026, Meta announced the Muse Wonder series — Muse Wonder and Muse Wonder Full — calling them its most capable models to date. They are Meta's first frontier-class models since its AI rebuild began fourteen months ago, trained at the company's gigawatt-scale Prometheus data center. Meta built them specifically for "personal and professional agentic work": handling insurance claims, understanding leases, making sense of finances, and frontier coding. Neither model ships with publicly downloadable weights — a decisive shift from the open-weight Llama era toward paid, controlled deployment.
Two Models, Two Ways In
- Muse Wonder: the fast frontier model, available now through the Meta Model API (model ID muse-wonder-1.0) and inside Muse Code; 1M-token context window; text, image, video, audio and PDF in, text out; API pricing reported at $4 per million input tokens and $20 per million output tokens — the same card as Claude Opus 5.5
- Muse Wonder Full: the stronger sibling, available ONLY to Muse Code subscribers ($50/month Power Usage or $200/month Maximum Usage plans), not served through the public API
| Model | Access | Reported price |
|---|---|---|
| Muse Wonder | Meta Model API + Muse Code | $4 / $20 per 1M tokens |
| Muse Wonder Full | Muse Code subscribers only | $50–$200/mo plans |
| Claude Opus 5.5 (comparison) | API | $4 / $20 per 1M tokens |
Active Memory and Active Focus
Meta is marketing two headline capabilities. Active Memory lets the model carry what matters across conversations and retrieve it when the task needs it — Meta tests this with an internal MemoryBench where both Wonder versions score higher than rival models. Active Focus is about persistence: when something fails, the model tries another way, works around the problem, asks for clarification when unsure, and checks in before consequential actions. Both target the two classic agent failure modes: losing context over long sessions and drifting away from the original goal.
“Introducing Muse Wonder, our new family of frontier models with state of the art performance for personal and professional agentic work.”
Where It Lands vs the Competition
On Meta's own 14-row scorecard, Wonder Full wins 7 rows (led by SWE-Marathon 57.5 vs Opus 5.5's 46.3 and DeepSWE 1.1 at 79.0 vs 74.2), while standard Wonder wins 5 and loses 9 against Opus 5.5. On the independent Artificial Analysis Intelligence Index, Wonder Full scores 56 and Wonder scores 52 — just behind Claude Opus 5.5 at 58, roughly level with Claude Sonnet 5.5 (56) and GPT-6 Astra (53). It is frontier-class, but at launch it trails the model it is priced identically to.
Frequently Asked Questions
What is the difference between Muse Wonder and Muse Wonder Full?
Muse Wonder is the API-served model ($4/$20 per 1M tokens via the Meta Model API and Muse Code). Muse Wonder Full is the stronger version that tops Meta's internal scorecard, but it is only available to Muse Code subscribers at $50–$200/month and is not served through the public API.
Are Muse Wonder weights open source?
No. Meta did not announce any open-weight release for either Muse Wonder or Muse Wonder Full. This is a deliberate shift from the Llama era — Meta's most capable models are now paid, controlled deployments only.
What are Active Memory and Active Focus?
Active Memory is Meta's mechanism for carrying relevant details across conversations and retrieving them when needed; Active Focus keeps the model working toward a stated goal, trying alternatives and asking for clarification before consequential actions. Both aim at long-running agent reliability.
How does Muse Wonder compare to Claude Opus 5.5?
On Meta's own scorecard Wonder Full beats Opus 5.5 on coding rows like SWE-Marathon and DeepSWE 1.1, but on the independent Artificial Analysis index Wonder Full (56) and Wonder (52) trail Opus 5.5 (58). Both are priced identically at $4/$20 per 1M tokens.
Related Prompts
Invisible Person Chair Editorial Poster Prompt — Conceptual Fashion Art
ChatGPT
Futuristic Sneaker Drop Poster Prompt — Limited Edition Campaign
ChatGPT
Oversized Puffer Jacket Fashion Editorial Poster Prompt — Winter Luxury
ChatGPT