Meta Releases Two New AI Models, With No Independent Test of Them Yet
Meta released two new artificial intelligence models, Muse Wonder and Muse Wonder Full, on Oct. 12, 2026, and made both available the same day. Meta's newsroom release says Muse Wonder went out to developers through the Meta Model API and the company's Muse Code product, and that Muse Wonder Full went to Muse Code subscribers.
Meta says the models are built for jobs that arrive with documents attached, among them working through an insurance claim, understanding a lease and making sense of someone's finances. The company also says they cover expert domains including law, finance and health along with coding.
The release names two behaviors as new. Active Memory carries details from one conversation into the next. Active Focus is meant to keep a model working when a task gets hard, try other approaches, recover from errors and ask for clarification before acting on anything important. Meta also states that the models deliver state-of-the-art performance on personal and professional agent tasks. No benchmark results accompany that wording in the release, and no independent evaluation of either model has been published.
On safety, Meta says it trained the Muse Wonder models to avoid sharing private information without permission, to turn down requests to help with harmful acts, and to keep working for developers and security teams who look for flaws in their own code. The company says it trains, evaluates and releases its most capable models in line with its Meta Superintelligence Scaling Framework, its own internal standard, and that a full report with additional safety evaluations will be available soon. That report has not been published.
Meta says the new models build on Muse Spark, which it released in April 2026, and that it began an overhaul of its AI work 14 months before this release. The release says Muse Wonder will soon power more of Muse, Meta's personal AI agent, reaching subscribers first.
