Meta released Muse Wonder and Muse Wonder Full, which it called "our most capable models to date, built to deliver state-of-the-art performance on personal and professional agentic tasks." They build on the groundwork of April's Muse Spark and were trained at Meta's gigawatt-scale Prometheus data center, the company said. The pitch rests on "Active Memory," carrying relevant context across conversations, and "Active Focus," staying with a hard task, trying other approaches, and checking in before important actions, aimed at knowledge work in law, finance, and health as well as coding. Per the model card, inputs are text, image, video, and audio, with a 1-million-token context window. Muse Wonder is available through the Meta Model API and Muse Code, Muse Wonder Full only to Muse Code subscribers, and the weights are not released. On Artificial Analysis's independent index, Full scored 56, fourth of 694 models, and Wonder 52. The same firm listed Wonder's API price at $4 per million input tokens and $20 per million output, and put the cost per index task at $4.52, nearly three times the $1.60 it had measured for Muse Spark 1.3. Meta also called it "our most aligned model to date" and said a full report with further safety evaluations would follow.
12October 2026
ModelConfirmed
Muse Wonder and Muse Wonder Full released, Meta's "most capable models to date"
Participants
Tags
- llm
- launch
- frontier-models
- agents
- pricing
Sources
- Introducing Muse Wonder and Muse Wonder Full: Frontier Models for Agentsen
- Muse Wonder: A Frontier Model For Personal & Professional AI Agentsen
- Muse Wonder Model Carden
- Muse Wonder (Max) - Intelligence, Performance & Price Analysis (Artificial Analysis)en
- Meta's new frontier model Muse Wonder Full breaks into the AI top tier (The Decoder)en
Later developments
23 October 2026 · Correction
Meta updated the model card, correcting Muse Wonder Full's ExploitGym results: after fixing its harness to include the intended vulnerability in the prompt, as the benchmark recommends, the model captured 20 intended flags (the original card had reported 92 flags in total, 21 of them intended). The original mention of 71 "unintended vulnerabilities" disappeared without explanation, the watchdog The Midas Project noted. The additional safety report promised at launch had not appeared.