Meta Muse Wonder Explained: Price, Access, and How It Compares

Quick Answer

• Meta launched Muse Wonder and Muse Wonder Full on October 12, 2026. TechCrunch calls them Meta’s first frontier models.
• Muse Wonder is available through the Meta Model API and Muse Code. Muse Wonder Full is for Muse Code subscribers only.
• API price for Wonder is $4 per million input tokens and $20 per million output tokens.
• Meta says the models match top rivals. An independent index (reported) still puts Claude Opus 5.5 ahead.
• Meta has not said whether the Wonder models will be open source.

What is Muse Wonder?

Muse Wonder is a new family of AI models from Meta’s Superintelligence Lab. There are two versions: Muse Wonder and Muse Wonder Full. They are bigger and more capable than Meta’s earlier lightweight Muse Spark models, with better results on hard tasks like coding.

Meta calls the launch “the next step on our scaling ladder” and “a major leap toward personal superintelligence.” This is the company’s first frontier release since it restructured its AI work in August 2025.

If you have used Meta’s tools before, see our reports on Meta Muse for Mac and Amazon blocking the Meta Muse agent.

Muse Wonder vs Muse Wonder Full

Meta also says Wonder will power future Muse assistant features, first for paid subscribers. Muse Code pricing was not stated in the sources we checked.

Muse Wonder price

TechCrunch says this is roughly on par with GPT-6 Sol and Claude Opus 5.5. The price is for the API version, not Wonder Full.

Key features

• Active memory: the model keeps and reuses what you tell it over a long time. Meta research scientist Kavya Kopparapu said the team worked to help the model “remember what’s important to you.”
• Active focus: the model keeps working through blockers on long tasks. This detail comes from ProPakistani’s coverage.
• Large context and many inputs: ProPakistani reports a 1 million token context window, with text, image, video, and audio input. TechCrunch did not state the context size, so treat this as reported by one outlet.
• Training: Meta says the models were trained at its Prometheus data center in central Ohio, which came online earlier in 2026.

How does it perform?

There are two kinds of numbers: Meta’s own and independent ones.

Meta’s claims. Meta’s model card says Wonder and Wonder Full score about the same as Claude Fable 5.1 and GPT-6 Astra. Meta says Wonder tops MemoryBench against Fable, Opus, and Astra.

Reported results by ProPakistani (from Meta’s tests):

Opus 5.5 also leads on HealthBench Pro, OSWorld (computer use), and Terminal-Bench in the same report.

Independent index. ProPakistani also lists the Artificial Analysis Intelligence Index:

So Wonder Full looks strong on long coding and document-heavy agent tasks in Meta’s testing, while Opus 5.5 still leads the main independent index. Independent testing is still early, and benchmarks do not show the full picture.

Is Muse Wonder open source?

Not as far as we know. TechCrunch says Wonder Full is not served through the API, and Meta has not said whether it will open any Wonder model. Meta did recently release the weights of a different model called Muse Glimmer.

How does it compare with Claude and Gemini?

Meta is now in the same price range as Claude Opus 5.5 and other top models. For how the others compare, read our guides on Claude Sonnet 5.5, Gemini 4 Argon, and ChatGPT vs Claude.

Should you try it?

• Developers: you can test Wonder through the Meta Model API. Compare it on your own tasks before you switch.
• Coders: Wonder Full needs a Muse Code subscription.
• Casual users: wait for the Muse assistant rollout, which starts with paid subscribers.

FAQs