Understand a repository before editing
Share the relevant files to compare implementation options and understand how a change will affect existing behavior.
Muse Wonder
Muse Wonder is Meta’s reasoning model for coding and complex tasks.
Muse Wonder
What is Muse Wonder AI?
Muse Wonder AI is a multimodal reasoning model from Meta, designed for complex software engineering and agent workflows.
- AI coding assistant
- The Muse Wonder model can explain unfamiliar code, propose changes and reason through errors when you share the relevant context.
- A million-token model context
- The upstream Muse model offers a 1,048,576-token context window; each app’s upload and conversation limits remain separate.
- Multimodal understanding
- Muse AI accepts text, images, video, audio and PDF inputs upstream, and returns text rather than generated images or video.
- Tool-ready reasoning
- Muse Wonder API supports tool calls and structured output, giving developers a way to connect reasoning with their own services.
- Configurable reasoning effort
- Developers can adjust reasoning effort through the provider API and balance the depth of an answer against its token cost.
- Chat in your browser
- Start a conversation, explore ideas and get help with code in Muse Wonder.
Why use Muse Wonder for coding?
Muse Wonder combines code understanding with reasoning, so developers can discuss a problem before choosing a change.
Turn errors into a testable explanation
Provide the error and expected behavior so the assistant can suggest a likely cause and a focused way to verify it.
Review changes against a concrete goal
Review a patch against its acceptance criteria and ask for edge cases, inconsistent behavior and useful validation.
Keep a human review step
Treat generated code as a proposal: read it, run suitable tests and check permissions before applying it in production.
How to use the AI coding assistant
Muse Wonder works best when your prompt includes the goal, relevant source and the result you expect to receive.
Describe your coding task
Sign in, describe the expected result and provide relevant source or logs while leaving credentials out of the prompt.
Give the model clear constraints
Choose the result you need: an explanation, patch, review or plan, with compatibility requirements and a clear scope.
Review and verify the answer
Review the answer, verify changes in your development environment and share the results when you need a revision.
Muse Wonder AI coding and reasoning
Muse Wonder AI coding support covers explanations, implementation plans and reviews without needing a terminal to start.
Muse Wonder AI coding assistant
Move from a feature idea to an implementation plan with dependencies and acceptance criteria your team can discuss.
Example prompt
Help me turn this feature idea into an implementation plan. List dependencies, tradeoffs and acceptance criteria.
Who is the Muse Wonder model for?
Muse Wonder helps developers, learners and technical teams reason through code and explain decisions in clear language.
Developers building an application
Discuss a feature, outline a change and review a patch while keeping the conversation tied to your users’ requirements.
Learners reading unfamiliar code
Request a plain-language explanation and a small example, then compare the answer with reliable learning material.
Teams evaluating an AI workflow
Evaluate representative tasks for correctness, useful output and token spend before choosing a model for team workflows.
Muse Wonder coding workflows and API access
Explore Muse Wonder models, coding workflows and API access.
| Muse Wonder for complex coding tasks | Muse Code versus browser chat | Muse Wonder API through OpenRouter |
|---|---|---|
| Discuss code, investigate errors and compare implementation approaches. | Muse Code runs a terminal agent; browser chat here does not automatically edit repositories or control your computer. | OpenRouter provides an OpenAI-compatible endpoint with token-based billing. |
Muse Wonder pricing and benchmarks
Muse Wonder pricing depends on usage, while benchmark results describe an evaluation setup rather than a guaranteed outcome.
Muse Wonder rates on OpenRouter
- Input
- $4
- Output
- $20
- Cached input
- $0.20
OpenRouter lists $4 input, $20 output and $0.20 cached input per million tokens, checked on October 13, 2026. These are upstream provider rates; this site's chat and API credit prices are shown in Pricing.
View OpenRouter ratesMuse Wonder benchmarks need context
Compare the edition, reasoning settings, tools and test setup before using a benchmark score to choose a coding model.
Muse Wonder subscription choices
Choose monthly or annual billing, or buy credits with a one-time pack.
Muse Wonder Pricing
Choose a subscription or a credit pack, and start with 10 trial credits; input and output usage both count.
Annual option
Pay once and receive your annual credit allowance up front.
Annual credits
Check the displayed allowance before checkout
Creator Yearly
$0.01 per credit
$14.50
$174 charged once a year
Works with
AI coding chat
Credit allowance
17,400 credits for the year, up front
Credit rate
$0.01 per credit on the yearly plan
17,400 credits, delivered up front
What's included
- Writing and code assistance
- Usage-based credits
- Account usage history
MOST POPULAR
Studio Yearly
$0.01 per credit
$24.50
$294 charged once a year
Works with
AI coding chat
Credit allowance
29,400 credits for the year, up front
Credit rate
$0.01 per credit on the yearly plan
29,400 credits, delivered up front
What's included
- Writing and code assistance
- Usage-based credits
- Account usage history
Max Yearly
$0.01 per credit
$49
$588 charged once a year
Multiplier
Works with
AI coding chat
Credit allowance
58,800 credits for the year, up front
Credit rate
$0.01 per credit on the yearly plan
58,800 credits, delivered up front
What's included
- Writing and code assistance
- Usage-based credits
- Account usage history
Pay safely and securely with
Also available at checkout in supported regions: EPS, Kakao Pay, Naver Pay, PAYCO, Samsung Pay, MB WAY and SEPA Direct Debit. Checkout shows the methods that work for your country, currency and plan.
Muse Wonder questions and answers
Muse Wonder answers below explain the model, API access and subscriptions so you can choose the right way to use it.
What is Muse Wonder?
Muse Wonder is Meta’s multimodal reasoning model for coding and complex agent tasks. It accepts several input types upstream and returns text.
Is Muse Wonder a coding model?
Yes. Complex software engineering is a key use case, alongside general reasoning and document understanding. Share the relevant code and requirements, then verify proposed changes with your own tests.
Where can I find Muse Wonder API access?
Create an API key in your account settings and use the OpenAI-compatible chat completions endpoint. See the API guide for authentication, examples and usage limits.
How does Muse Wonder pricing work?
Choose a monthly or annual subscription, or a one-time credit pack. Credits cover chat and API usage; the pricing page lists each plan’s price and credit allowance.
What affects Muse Wonder cost?
Prompt length, conversation history, reasoning effort and response length affect token use. Long context can improve understanding but also increases spend, so give the model relevant material rather than an entire unrelated repository.
How should I read Muse Wonder benchmarks?
Check which edition was tested, the reasoning settings and the tools available during evaluation. Use a small set of your own coding tasks to compare results before making a decision based on a published score.
Does Muse Wonder generate images or videos?
No. Muse Wonder returns text and code. It can help explain, analyze and reason about the material you provide.
Do I need Muse Code or OpenCode to start?
No. You can begin with a browser conversation here. Muse Code and OpenCode are separate agent tools for development environments, and their integrations, permissions and billing depend on their own setup.
How should I use Muse Wonder Reddit discussions?
Community discussions can offer examples and impressions from particular workflows. Check claims about availability, pricing and model capabilities against the current Meta documentation and provider model page, then test the model on your own representative task.