Muse Wonder: what it is and what it costs

Muse Wonder is Meta's first frontier-class AI model family, released on October 12, 2026, in two versions: Muse Wonder and the larger Muse Wonder Full. The standard model costs $4 per million input tokens and $20 per million output tokens on the Meta Model API, and Muse Wonder Full is only available inside a paid Muse Code subscription.

What Meta posted on X

Mark Zuckerberg announced it himself. In his launch post, @finkd noted that Meta Superintelligence Labs is just over a year old. He described Muse Spark as the lab's first model family and Muse Wonder as "a much larger and more advanced class of models that brings us to the frontier." Muse Wonder is the fast version. Muse Wonder Full is the full-size model.

The official account, @AIatMeta, covered availability. Muse Wonder is live now through the Meta Model API and Muse Code, Meta's coding tool. "Muse Wonder Full is available today for Muse Code subscribers," the post says. Meta also says Muse Wonder will "soon" power its Muse personal agent app, starting with subscribers. Meta pitches both models at personal and professional agentic tasks: jobs where the model works through many steps on its own instead of answering a single question.

Muse Wonder vs Muse Wonder Full

The difference that matters most is access:

• Muse Wonder is on the API at $4 input and $20 output per million tokens, according to TechCrunch, which puts it roughly in line with GPT-6 Sol and Claude Opus 5.5. Per Artificial Analysis, it has a 1 million token context window and takes text, images, video and audio as input.
• Muse Wonder Full is not on the API. Meta's Muse Code page lists it on the Power Usage plan at $50 a month and includes it in the $200 Maximum Usage plan.

So if you're building something on Meta's API, you get the smaller model. To use the bigger one, you have to work inside Meta's coding app.

Meta's launch post leans on two features. Active Memory carries important context about you from one conversation to the next and pulls up the details that matter for the current task. Active Focus keeps working on a hard goal: it tries other approaches, works around blockers, asks when something is unclear and checks in before important actions. The launch post doesn't mention open weights, which used to be what set Meta's models apart.

What the independent numbers say

@ArtificialAnlys, the benchmarking group Artificial Analysis, posted scores within hours. On its Intelligence Index v4.3 at max reasoning effort, Muse Wonder scored 52 and Muse Wonder Full scored 56, up from 48 for Muse Spark 1.3, which came out about six weeks earlier. The biggest gains were on agentic tasks, including coding agents. The group also noted that running Muse Wonder through its test suite cost more than many comparable models.

Meta's own model card is more ambitious. According to TechCrunch, Meta says both models match Claude Fable 5.1 and GPT-6 Astra, and that Muse Wonder beats them on MemoryBench. MemoryBench is Meta's own internal test, so take that one carefully until outside testers have checked it.

My read

Meta now has a model in the frontier conversation, and that's good news for anyone paying for AI, since more competition at the top tends to push prices down. The launch-day picture is mixed, though. The model you can build on, the API version, scores below the Full model on the independent index, and the strongest memory claim comes from a test Meta wrote itself.

The memory pitch is the part to watch. If Active Memory holds up, an assistant that remembers a customer's history across weeks of conversations is useful for service businesses. I'd test that against your own customer conversations before trusting a benchmark chart on it.

What this means for businesses using AI

A new frontier model every few weeks is normal now. The businesses that benefit from it are the ones whose automations don't depend on a single vendor, so trying Muse Wonder on a real task means changing a setting instead of rebuilding a workflow. Whether the model you're on is the newest one matters less. If you'd like a second set of eyes on which of your processes are ready for that kind of setup, New Face Design offers a free process audit.