Meta Releases Muse Wonder, Its First Frontier AI Model, Built Around Memory
The model is on Meta's API at $4 per million input tokens, but its top version is limited to Muse Code subscribers.
Meta's Superintelligence Lab launched Muse Wonder and Muse Wonder Full on Monday, October 12, according to TechCrunch. TechCrunch calls them Meta's first frontier-class models, larger and more capable than the earlier Muse Spark, especially on complex work like coding. Both will power Meta's Muse agent.
Meta stresses "active memory," meaning the model keeps and uses what you tell it over long periods. That matters for assistants that help with the same work for weeks. TechCrunch says Meta's own tests show Muse Wonder scoring higher than Fable, Opus or GPT-6 Astra on MemoryBench, which tests memory building. Those results come from Meta, not independent testers.
Muse Wonder is available through the Meta Model API at $4 per million input tokens and $20 per million output tokens, which Meta says is roughly in line with GPT-6 Sol and Opus 5.5. Muse Wonder Full is limited to Muse Code subscribers and is not offered through the API.
You do not need to switch tools today. If you pay for an AI assistant mainly for memory over long projects, watch for independent tests before trusting Meta's claims. If you use apps built on AI models, expect more of them to offer Meta as an option.
- Meta releases its first frontier model, Muse Wonder (TechCrunch)
Is AI recommending your business?
Our AI visibility audit checks what ChatGPT, Gemini and Google's AI say about you, and includes a 1:1 review of the results.
See the audits