Meta’s Muse Wonder Is a Frontier Model. The API Price Tells You Who It Is For

On 12 October 2026, Meta’s Superintelligence Lab released Muse Wonder and Muse Wonder Full. TechCrunch described them as Meta’s first frontier models: larger than Muse Spark, aimed at harder work such as coding, and priced like the other models at the top of the market. The release matters less as a leaderboard and more as a bill you might pay.

The facts in the launch

Muse Spark was the lighter, lower-latency model. Wonder is the step up. Meta’s own tests, as reported from the model card, put Wonder and Wonder Full near Fable 5.1 and GPT-6 Astra. Through the Meta Model API, Wonder is $4 per million input tokens and $20 per million output tokens, in the same range TechCrunch cited for GPT-6 Sol and Opus 5.5. Wonder Full is for Muse Code subscribers and is not on the API. Both models also power the Muse agent. Meta called the pair a step toward personal superintelligence. Report: https://techcrunch.com/2026/10/12/meta-releases-its-first-frontier-model-muse-wonder/

How to read the price

• Output is the expensive line. A long answer costs five times the input rate. Short, structured replies protect the budget.
• Wonder Full is a product, not an API. If your app needs that tier, you are depending on Muse Code, not on a key you control.
• Benchmarks are Meta’s. Ask for your prompts: a quotation, a support reply in your tone, a code change in your repo.
• The Muse agent is the distribution. A model you never call directly can still be what a shopper’s agent uses. See what Meta Muse is.

Most business workflows do not need a frontier model on every turn. Use a cheaper model to classify and retrieve. Call Wonder, or whoever leads your eval, when the task is long, ambiguous, or expensive to get wrong. That split is how AI agent development stays affordable after the demo.

FAQ

Is Muse Wonder the same product as the Muse app?

No. Wonder is a model. The Muse agent is the product people set up. Meta said these models will power that agent. Your integration choice is still API, Muse Code, or your own stack.

Should we budget Wonder for an India launch?

Budget the tokens you will actually send. A support bot that pastes a whole policy into every prompt will look cheap in a pilot and expensive in a month. This note is coverage of a launch, not a recommendation to buy Meta stock.