About

Meta launches Muse Wonder and open-weight Muse Spark Open; APAC small firms report 84% AI use

Published
Score
16

Why it matters

Meta launched two proprietary frontier models on Monday, October 12, 2026: Muse Wonder and Muse Wonder Full, both from Meta Superintelligence Labs, the research group led by former Scale AI CEO Alexandr Wang. Neither model will be released with publicly downloadable weights. Muse Wonder reportedly carries a monthly subscription of $20 to $100. Three days later, Meta released Muse Spark Open, an open-weight model with 816 billion total parameters, 42 billion active per token, and a 1-million-token context window. Meta says it can run on two Apple Mac Studio computers rather than in a data center, and it describes the model as the "most capable US open-weight model to date." The model launched October 15 across seven inference clouds and four weight hosts. Separately, Meta published a YouGov survey it commissioned of 3,983 small business owners in nine Asia-Pacific markets. It found 84% use AI, 76% say AI has improved productivity, and 86% say small businesses need more guidance to benefit from it.

Several points remain open. The survey findings were commissioned and published by Meta, and its October 14 publication date falls one day after the story date, so the timing needs confirmation before it is cited. The survey's 86% figure shows unmet demand for guidance, not a measured rise in business worry, so the "growing concern" framing is not supported by the data. Muse Wonder's pricing is reported, not confirmed in Meta's own materials. A Decrypt report says Meta employees consumed 60 trillion AI tokens over 30 days earlier this year, an estimated $900 million at list prices. The amount Meta actually paid was likely lower, and that figure has not been disclosed. Meta has also said it will keep spending despite calls for a coordinated slowdown, with 2026 capital expenditures planned at $130 billion to $145 billion.

For attorneys, the near-term items are contract and cost terms. Clients weighing AI subscriptions should compare list prices with negotiated enterprise rates, and any vendor agreement should address data use, model access, and what happens if pricing changes. The open-weight release matters for licensing review, since Muse Spark Open's license terms govern how clients can deploy and modify it, and that shift away from Meta's earlier Llama approach changes the analysis. Watch for the full license text, any regulatory response to Meta's spending or to competition with OpenAI and Anthropic, and whether the small business survey gets cited in policy debates about AI guidance.

Sources

mail Subscribe to Artificial Intelligence email updates

Primary sources. No fluff. Straight to your inbox.

Also on LawSnap