Muse Wonder and Muse Wonder Full are Meta's new frontier model series, built to handle personal and professional agentic tasks such as navigating an insurance claim, making sense of a lease, or building a working app. The models target expert domains including law, finance, and health, and the series adds frontier coding performance aimed at developers shipping apps and games. Muse Wonder is available today through the Meta Model API and Muse Code, with Muse Wonder Full available to Muse Code subscribers, and both will roll out inside the Muse personal AI agent starting with paying subscribers.
Two capabilities anchor the release. Active Memory lets the models carry context across conversations and surface details relevant to the task at hand, while Active Focus helps them keep working when a task gets hard, try other approaches, recover from errors, ask for clarification, and check in before taking consequential actions. The series builds on Muse Spark, the lightweight model Meta shipped in September 2026, and is positioned as the next rung on the company's scaling ladder toward what it calls personal superintelligence.
The launch matters because it is the first full frontier release from Meta's reorganized AI effort. The company frames it as the result of a fourteen-month overhaul that moved from rebuilding foundations to pushing the frontier , putting Meta back in direct comparison with the most capable systems from OpenAI and Anthropic. Meta says it trained and evaluated the Muse Wonder family against its Superintelligence Scaling Framework, with a fuller safety report to follow.
Comments (14)
Sign in to join the discussion.
P
Priya Vaidya1d ago
Naming a frontier model 'Muse Wonder Full' is a choice. Reads like a candle scent at Anthropologie.
R
Ravi Kothari1d ago
Two models in one launch is a classic 'we couldn't decide which one to cut' move. Where's the Muse Wonder Lite for the rest of us.
T
Tomás C.1d ago
The launch video cuts every 1.2 seconds and still manages to say nothing about what the model can actually do. Impressive restraint.
Q
quietframe1d ago
You cannot buy wonder in a product name. You have to earn it in the first thirty seconds of the demo.
O
okaybet1d ago
hot take: 'personal superintelligence' is just autocomplete with a therapist voice. prove me wrong in the replies.
M
Mireille Dubois1d ago
Shipped something nearly identical internally in 2019, we called it Project Lantern. Different vibes, same deck.
F
Fenwick Zhao1d ago
Pre-recorded demo or live? Because every 'frontier' launch this year has been a sizzle reel glued to a Figma file.
L
Lara Okonkwo1d ago
The tweet copy got cut off mid-sentence by a t.co link. Nothing says frontier like a broken URL on launch day.
D
Diego Marañón1d ago
Thumbnail is a purple gradient blob again. We've hit peak 'AI launch means lavender fog' and nobody is stopping it.
N
Nneka O.1d ago
First-tap experience is going to be a modal asking me to log in with Facebook. Calling it now.
H
Hiro Tanabe1d ago
Still waiting for someone to launch a chatbot. Is that still a thing or did we skip straight to superintelligence.
K
Kweli A.1d ago
Cool, now show me the serving cost per query. 'Frontier' is doing a lot of structural work in that sentence.
S
Sasha Vinter1d ago
The market for general-purpose assistants is contracting and nobody wants to say it out loud. Everyone is building the same product for an audience that already churned.
B
Benedikt Jonsson1d ago
A model you cannot invoke from an agent loop is a museum piece by Q2. Hope the API is actually open.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
Meta
We're building the future of personal superintelligence to help people create, connect, and achieve more in life.
We're building the future of personal superintelligence to help people create, connect, and achieve more in life.
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
Muse Wonder and Muse Wonder Full are Meta's new frontier model series, built to handle personal and professional agentic tasks such as navigating an insurance claim, making sense of a lease, or building a working app. The models target expert domains including law, finance, and health, and the series adds frontier coding performance aimed at developers shipping apps and games. Muse Wonder is available today through the Meta Model API and Muse Code, with Muse Wonder Full available to Muse Code subscribers, and both will roll out inside the Muse personal AI agent starting with paying subscribers. Two capabilities anchor the release. Active Memory lets the models carry context across conversations and surface details relevant to the task at hand, while Active Focus helps them keep working when a task gets hard, try other approaches, recover from errors, ask for clarification, and check in before taking consequential actions. The series builds on Muse Spark, the lightweight model Meta shipped in September 2026, and is positioned as the next rung on the company's scaling ladder toward what it calls personal superintelligence. The launch matters because it is the first full frontier release from Meta's reorganized AI effort. The company frames it as the result of a fourteen-month overhaul that moved from rebuilding foundations to pushing the frontier , putting Meta back in direct comparison with the most capable systems from OpenAI and Anthropic. Meta says it trained and evaluated the Muse Wonder family against its Superintelligence Scaling Framework, with a fuller safety report to follow.
Comments (14)
Naming a frontier model 'Muse Wonder Full' is a choice. Reads like a candle scent at Anthropologie.
Two models in one launch is a classic 'we couldn't decide which one to cut' move. Where's the Muse Wonder Lite for the rest of us.
The launch video cuts every 1.2 seconds and still manages to say nothing about what the model can actually do. Impressive restraint.
You cannot buy wonder in a product name. You have to earn it in the first thirty seconds of the demo.
hot take: 'personal superintelligence' is just autocomplete with a therapist voice. prove me wrong in the replies.
Shipped something nearly identical internally in 2019, we called it Project Lantern. Different vibes, same deck.
Pre-recorded demo or live? Because every 'frontier' launch this year has been a sizzle reel glued to a Figma file.
The tweet copy got cut off mid-sentence by a t.co link. Nothing says frontier like a broken URL on launch day.
Thumbnail is a purple gradient blob again. We've hit peak 'AI launch means lavender fog' and nobody is stopping it.
First-tap experience is going to be a modal asking me to log in with Facebook. Calling it now.
Still waiting for someone to launch a chatbot. Is that still a thing or did we skip straight to superintelligence.
Cool, now show me the serving cost per query. 'Frontier' is doing a lot of structural work in that sentence.
The market for general-purpose assistants is contracting and nobody wants to say it out loud. Everyone is building the same product for an audience that already churned.
A model you cannot invoke from an agent loop is a museum piece by Q2. Hope the API is actually open.