Meta shipped Muse Wonder and Muse Wonder Full on Monday, its first frontier-class models, closing a fourteen-month rebuild of the AI unit under Alexandr Wang. The models are live inside the Meta Model API and Muse Code today; the Muse consumer agent will inherit them next, behind a $20 or $100 monthly subscription. API pricing lands at $4 per million input tokens and $20 per million output tokens, per TechCrunch.
For most readers, the interesting beat isn’t the pricing sheet. It’s what sits downstream of the model switch.
Meta’s leadership is framing Wonder as on par with OpenAI’s ChatGPT and Anthropic’s Claude on coding and on financial and legal questions, according to Bloomberg. The parity claim arrived without published benchmarks, as The Next Web noted, and Meta hasn’t said whether the weights will be released, a decision that shapes obligations under Article 53 of the EU AI Act. Call that posture what it’s: narrative management ahead of evaluation.
The two headline capabilities are more concrete. Active Memory keeps context across conversations, and Active Focus trains the model to work through setbacks instead of pinging the user at the first obstacle. “a tremendous amount of effort into making sure your model can remember what’s important to you over days, weeks, months, and beyond,” said Kavya Kopparapu, a research scientist at Meta, in a video accompanying the launch. Wang put the second capability in operational terms to CNBC: “When you do real-world tasks, you’re going to get roadblocks and you’re going to get things that come up or various setbacks. We specifically tried to train the model to be able to try other approaches instead of stalling and then checking back in with the user.”
Both models will power the Muse agent, per TechCrunch. That’s the part small B2C operators should mark. The Muse agent is where customer-facing conversations will increasingly get handled. A model that remembers what a shopper browsed last week, and that doesn’t stall halfway through a multi-step checkout, is a different conversion instrument than what shipped a quarter ago.
The caveats are real. Muse Wonder isn’t powering the Business Agent tier yet; the parity claims are Meta’s own; the Muse app only recently reached the top of Apple’s App Store, which tells you about distribution, not quality. Models were trained at Meta’s Prometheus data center in central Ohio, a detail that mostly signals how much capex is now sunk into the roadmap.
The practical move for owner-founders running inbound DMs: pull your highest-volume scripts, baseline their close rate this week, and re-run the same conversations once Wonder reaches the Business Agent. We walked through the setup in our earlier piece on Meta turning Muse into a customer-hunting tool for small businesses. The engine is being swapped underneath a product millions of businesses already use. That’s where the delta will show up first, and quietly.
Sources
- https://about.fb.com/news/2026/10/introducing-muse-wonder-models/
- https://techcrunch.com/2026/10/12/meta-releases-its-first-frontier-model-muse-wonder/
- https://www.bloomberg.com/news/articles/2026-10-12/meta-says-new-muse-model-reaches-ai-frontier-after-14-month-push
- https://www.cnbc.com/2026/10/12/meta-muse-wonder-ai-model-agent.html
- https://thenextweb.com/news/meta-muse-wonder-frontier-eu