Meta Muse Wonder: what it is and how to try it today

What Meta launched on October 12

On October 12, 2026 Meta introduced the Muse Wonder series: Muse Wonder and Muse Wonder Full. Meta calls them its most capable models to date, built for personal and professional agentic tasks: work that takes many steps, not just one answer. They build on Muse Spark, the model behind the Muse personal agent Meta launched in September.

• Muse Wonder is available today through the Meta Model API and Muse Code, Meta's coding agent for the terminal.
• Muse Wonder Full, the higher-capability variant, is available today only to Muse Code subscribers.
• Meta says Muse Wonder will soon power more of the Muse app, rolling out first to subscribers.
• Meta's docs list Muse Wonder as a preview.

Meta's own examples of what it is for: navigating an insurance claim, understanding a lease, making sense of someone's finances, and building a game or an app for a new business.

Active Memory and Active Focus, in plain English

• Active Memory: it carries what matters to you across conversations and brings back the details that are relevant to the task in front of it. You stop re-explaining your business every chat.
• Active Focus: it keeps working toward the goal when a task gets hard, tries other approaches, recovers from errors, asks for clarification when needed, and checks in before following through on anything important.

Meta also says it trained the models to avoid sharing private information without permission and to refuse requests to help with harmful acts. A full safety report is listed as coming soon.

Three ways to use it

1. Meta Model API (pay-as-you-go). Model ID muse-wonder-1.0. You pay per token, with no minimum. Good for a quick test or for plugging it into your own tools.
2. Muse Code. Meta's ready-made agent at the command line. Muse Wonder works on pay-as-you-go and on every subscription plan. Muse Wonder Full is only on the Power Usage and Maximum Usage plans.
3. The Muse app. Meta's personal agent on iOS, Android and muse.ai, rolling out in the US. Muse Wonder is coming to it soon, subscribers first. Until then the app runs on Muse Spark.

Meta notes that model availability can vary by country.

Try it today in Muse Code

1. Install Muse Code with Meta's one-line installer (macOS or Linux):

   curl -fsSL https://dev.meta.ai/install.sh | sh

   On Windows PowerShell: irm https://dev.meta.ai/install.ps1 | iex
2. Check it worked: muse --version
3. Open a folder (for example a copy of the documents you want to test) and run muse. Muse Code asks whether to trust the folder, then asks you to sign in with your browser or paste a Meta API key. You need a Meta account and a Meta Model API account; pay-as-you-go needs a payment method on the Billing page.
4. The default model is muse-spark-1.2, so switch: type /models and pick Muse Wonder, or start with muse --model muse-wonder-1.0.
5. Optional: set how hard it thinks with /effort (from minimal up to max).
6. Ask for your task in plain English, then approve its actions as it goes.

Privacy check before you start: in Standard mode Meta says your content is not used to train its models. Muse Code subscriptions default to Contributor mode, which gives you more usage in exchange for letting Meta train on your prompts, files and responses. You can turn it off any time in settings. On pay-as-you-go, Muse Wonder runs in Standard mode only.

What it costs

Meta Model API pay-as-you-go prices for muse-wonder-1.0 (Standard mode, per 1 million tokens):

• No long-context premium: the same rate whether the context is nearly empty or nearly full.
• Web search grounding adds $2.50 per 1,000 searches.
• A token is roughly 3–4 characters of English text.
• Context window: 1,048,576 tokens. It takes text, image, video, audio and PDF input.

Muse Code also has four flat-rate plans (Everyday, High, Power and Maximum Usage). Meta's docs say benefits can vary by region and are shown during sign-up, so check the price there before you subscribe. Subscriptions are managed in Meta's Accounts Center.

A first test for your business

Pick one real job you already do by hand, then compare the result with what you would have written yourself. For example, with a lease or service contract in the folder:

Read lease.pdf. List every date, fee and renewal deadline in a table with the page number for each. Then list anything that could cost me money that I should ask a lawyer about. Do not guess: if something is unclear, say so.

Or with a folder of invoices:

Read every invoice in this folder. Make a table of client, invoice number, amount, date and status. Flag anything overdue more than 30 days and draft a short, polite reminder for each one. Do not send anything.

The free prompt pack below has eight more like these.

Limits to know

• Muse Wonder is a preview model in Meta's docs.
• Muse Wonder Full is not on the API, only in Muse Code on the Power Usage and Maximum Usage plans. Setting reasoning effort to max on Muse Wonder does not select it.
• The Muse app does not have Muse Wonder yet: Meta says it is coming soon, subscribers first.
• Benchmark results are Meta's own testing.
• It is an AI. Check anything legal, financial or medical with a professional before you act on it.

Questions people also ask

Sources

• Meta Newsroom: Introducing Muse Wonder and Muse Wonder Full (October 12, 2026)
• Meta AI research blog: Muse Wonder
• Meta Model API docs: Models
• Meta Model API docs: Pricing and rate limits
• Meta Model API docs: Muse Code
• Meta Model API docs: Muse Code authentication and billing
• Meta Model API docs: Muse Code subscriptions
• Meta Model API docs: Muse Code configuration (model selection)
• Meta Newsroom: Introducing Muse (September 8, 2026)