Meta releases Muse Wonder and Wonder Full AI models
By Automate Basics, written with AI from TechCrunch's original
Meta has released Muse Wonder and Muse Wonder Full, two models aimed at more complex work than Muse Spark. Wonder is available through Meta's Model API, while Wonder Full is limited to Muse Code subscribers, according to TechCrunch.
What changed with Meta's Muse models?
Meta has released Muse Wonder and Muse Wonder Full, its first models intended to compete with leading models on demanding tasks, TechCrunch reports. Meta says the new models are larger and more capable than Muse Spark, which was designed to respond quickly with less computing power. Coding is one area where Meta reports better results.
Meta is also emphasizing what it calls active memory: the ability to retain and bring together information a user has provided over time. According to TechCrunch, Meta's own tests put Muse Wonder ahead of several competing models on a memory benchmark. Those results are company-reported scores, not a promise that the model will remember every detail of a work project.
Who can use Muse Wonder and Wonder Full?
Muse Wonder is available through the Meta Model API, TechCrunch reports. Meta prices information sent to Wonder at $4.00 for each million input tokens. Information it generates costs $20.00 for each million output tokens. Tokens are units used to measure the text a model processes.
Muse Wonder Full has a different access route: Meta says it is available only to Muse Code subscribers and will not be offered through the API. Both models will also power the Muse agent. The report does not give a Muse Code subscription price or spell out which model an individual Muse agent user will encounter. For a team weighing the options, the distinction is between paying for Wonder's API usage and accessing Wonder Full through a subscription.
How do I try Muse Wonder at work?
Muse Wonder is worth considering for a task that involves several pieces of project information, such as bringing together successive briefs and identifying what changed. That would put Meta's memory claim to a practical test without assuming the model can reliably recall everything from earlier work. TechCrunch reports that Wonder can be accessed through Meta's Model API; it does not provide a setup guide for workplace users.
Check any resulting summary against the original material, especially dates, decisions and details that were revised. If your work involves coding, a complex coding task is another relevant trial because Meta reports improved results in that area. Check the output against the task requirements rather than treating Meta's benchmark comparisons as a measure of performance on your own work.
Frequently asked questions
Can I use Muse Wonder Full through Meta's API?
No. TechCrunch reports that Meta is offering Muse Wonder Full only to Muse Code subscribers, not through the Meta Model API. Muse Wonder is the model available through the API. The report does not state the Muse Code subscription price.
What does Meta say Muse Wonder's memory can do?
Meta says Muse Wonder is designed to retain and combine information users provide over time. TechCrunch reports that Meta's tests gave it a strong score on a memory benchmark. The report does not establish how reliably it will remember the details of a particular workplace project.
How much does Muse Wonder cost through the API?
TechCrunch reports Meta's price as $4.00 per million input tokens, which measure information sent to Wonder, and $20.00 per million output tokens, which measure what it generates. That pricing applies to Muse Wonder through the API, not to Muse Wonder Full.
Related reading
Blueprints AI plans wider release of building design softwareBlueprints AI plans to make a new version of its building design software publicly available this week, TechCrunch reports. It can revise drawings across multiple disciplines when a project changes, but licensed professionals must review and seal documents before permit submission.
Honeyjar brings AI drafting and research to public relations workHoneyjar is building AI tools to help people plan and carry out public relations work, from drafting announcements to finding reporters. TechCrunch reports that the product has a waitlist and is being used by beta testers.
Google Meet adds routine follow-ups to its post-meeting screenGoogle Meet now offers routine next steps after a meeting, including drafting a follow-up email and adding action items to Google Tasks. Google says eligible Workspace users can use them when Take notes for me was enabled during the meeting.
Tools in this piece
Written with AI from TechCrunch's original and published after automatic checks: every figure here appears in the original, and no sentence is copied from it. The picture is AI-generated. The original is the authority.
Source: Meta releases its first frontier model, Muse Wonder, TechCrunch, 12 Oct 2026.