Meta announced two new AI models on October 12, 2026: Muse Wonder and a larger version called Muse Wonder Full. Both models scored in the top tier on the Artificial Analysis Intelligence Index, with Muse Wonder Full reaching 56 points and Muse Wonder 52 points. The scores place them among the leading frontier models released by other companies.
How the models differ
Muse Wonder Full adds two features Meta calls "Active Memory"—the ability to keep key facts across a conversation—and "Active Focus," which lets the model stick with hard problems, try different approaches, and ask follow‑up questions. Both models can process up to one million tokens and accept text, images, video, and audio.
The full version is currently only available to users of Meta's Muse Code platform, while the smaller Muse Wonder will soon be offered through the Muse app for paying subscribers and through Meta's API.
Performance and price
On the Terminal‑Bench 4.0 benchmark, Muse Wonder Full jumped from 33 % to 57 % success, landing in the top three on six of ten tests. Coding scores also rose, with Muse Wonder Full hitting 68 points on the Coding Agents Index, matching top models from Anthropic and others. However, factual knowledge remains lower than rivals, scoring only 30 points on the AA‑Omniscience test.
Costs have risen sharply. According to Artificial Analysis, running Muse Wonder costs $4.52 per task on the Intelligence Index, about three times the $1.60 cost for the previous Muse Spark 1.3. Token pricing is now $4 per million input tokens and $20 per million output tokens, compared with $1.25 and $4.25 previously.
Why it matters
For developers and businesses, the new Muse models offer stronger coding ability and better task persistence, but they come with higher per‑task and token fees. Users will need to weigh the performance gains against the increased cost, especially if they run many large‑scale prompts.
