Meta's Muse Wonder: What the New AI Model Means for Businesses and Developers

On 12 October 2026, Meta introduced Muse Wonder and Muse Wonder Full, which it calls its most capable AI models to date. Meta says they are built for personal and professional agentic tasks, from making sense of a lease or an insurance claim to building a working app. You can read the details in Meta's newsroom announcement and the longer Meta AI research post.

This is a different story from Meta's recent advertising updates. Muse Wonder is a general AI model that developers can build on and that will power Meta's Muse personal agent app. So in this post we explain what the models do, how you can access them, what they cost and what businesses should think about before building on them.

What Meta announced

Meta says it began a ground-up overhaul of its AI work 14 months ago and that Muse Wonder builds on its earlier Muse Spark models. There are two versions:

• Muse Wonder: available now to developers through the Meta Model API and in Muse Code, Meta's coding agent.

• Muse Wonder Full: a more powerful version, available now to Muse Code subscribers.

• Muse app: Meta says Muse Wonder will soon power experiences in the Muse app, rolling out first to subscribers.

What Muse Wonder can do, according to Meta

Meta highlights four areas. Each description below reflects Meta's claims:

• Domain expertise: handling problems in law, finance and health, such as understanding a lease or financial documents.

• Active Memory: carrying what matters to you across conversations and bringing back relevant details when a task needs them.

• Active Focus: continuing to work toward a goal when a task gets hard, trying other approaches and checking in before important actions.

• Frontier coding: supporting software projects from small apps to complex codebases.

Meta says the models achieve state-of-the-art results on several benchmarks for agents, memory and coding. These are Meta's own results. Independent testing was not available when we wrote this, and we have not tested the models ourselves.

Muse Wonder access and pricing

The table combines Meta's Model API pricing page, checked on 13 October 2026, with subscription prices reported by CNBC. Prices are in US dollars.

Access option	Price	Muse Wonder status
Meta Model API (pay as you go)	$4 per 1M input tokens, $20 per 1M output tokens, $0.20 per 1M cached input tokens	Available now as muse-wonder-1.0
Muse Code subscription	$50 or $200 a month, as reported by CNBC	Muse Wonder available now; Muse Wonder Full for subscribers
Muse app subscription	$20 or $100 a month, as reported by CNBC	Coming soon, subscribers first, according to Meta

A few more details from Meta's pricing page are worth knowing. Web search grounding costs $2.50 per 1,000 searches on top of token costs. There is no long-context premium. Meta's cheaper Contributor mode, where you let Meta use your content to train future models, is not available for Muse Wonder. In Standard mode, Meta says your content is not used to train its models.

In CNBC's interview, Meta AI chief Alexandr Wang said the API offering will cost more than Meta's standard subscriptions because it is a much larger, frontier-level model, but will still be priced "meaningfully below" leading models from OpenAI and Anthropic. That comparison is Meta's claim.

How Muse Wonder fits Meta's wider AI plans

Muse Wonder arrives a few weeks after Meta launched Muse, its personal AI agent app. CNBC reports that the app shot to the top of Apple's App Store after launch and that it is free, with paid plans for more usage. By putting its most capable model behind subscriptions and a paid API, Meta is testing whether people and developers will pay for its AI directly, rather than relying only on advertising revenue.

Meta's research blog also says Muse Wonder will power artifact generation and other processes that need more intelligence inside Muse, and that it plans to use Muse Wonder to develop specialised models it can bring to everyone using the app. For businesses, that suggests Meta's consumer AI features will keep improving quickly, which could shape how your customers search for and contact businesses on Meta's platforms.

What this means for businesses

This section is our view as a digital and app development agency, not Meta's claims.

More choice for building AI features

For businesses building AI into their own products, another strong model with published API pricing means more choice. A startup building a customer support assistant or a document review tool can now compare Muse Wonder with other providers on cost, quality and data terms. Competition usually helps buyers.

Memory and agents raise the stakes on privacy

Meta's focus on memory and agents that take action is useful, but it also means these systems may handle messages, calendars and documents. CNBC reports that some Muse users complained the agent accessed personal messages without authorisation, which Meta disputes. Whatever the cause, it is a reminder that any AI agent you deploy needs clear permissions and simple ways for users to see what it can access.

Test on your own tasks

Benchmark wins don't guarantee results on your workflow. If you are considering Muse Wonder for a product feature, run the same small set of real tasks through two or three models and compare accuracy, speed and cost before you commit.

Limitations to keep in mind

• Safety report pending: Meta says a full report with additional safety evaluations will be available soon.

• Muse app rollout not complete: Muse Wonder in the Muse app is described as coming soon.

• Subscription prices are reported, not listed by Meta in its announcement: confirm current plans in the app before you subscribe.

• Usage limits: Meta's API page lists 3,000 requests and 4 million tokens per minute per team for Muse Wonder.

• Benchmarks are self-reported: wait for independent tests or run your own.

Our take: should your business use Muse Wonder?

1. If you only use AI for writing and research, there is no rush. Keep using the tools your team already knows.

2. If you are building an AI feature into an app or website, add Muse Wonder to your shortlist and test it on real tasks.

3. Estimate costs early. Output tokens cost five times more than input tokens on Meta's API, so long answers add up.

4. Decide what data an agent may access before you connect it to email, files or customer records.

Planning an app or website feature that uses AI? Our app development team can help you compare models and build it properly, or you can book a consultation to talk it through. For no-code options, see our list of AI tools to automate and grow your business.

FAQ

Is Muse Wonder free?

No free access was announced. Developers pay per token through the Meta Model API, and CNBC reports that Muse app users will need a $20 or $100 monthly subscription to use it.

What is the difference between Muse Wonder and Muse Wonder Full?

Meta describes Muse Wonder Full as the more capable version. At launch it is available only to Muse Code subscribers, while Muse Wonder is also available through the Meta Model API.

Does Meta train on my API data?

Meta's pricing page says that in Standard mode your content is not used to train its models. Contributor mode, which offers discounted pricing or more usage in exchange for permission to train on your content, is not offered for Muse Wonder.

---

Sources

• Meta Newsroom, "Introducing Muse Wonder and Muse Wonder Full: Frontier Models for Agents", 12 Oct 2026

• Meta AI Research, "Muse Wonder", 12 Oct 2026

• Meta Model API, "Pricing and rate limits", checked 13 Oct 2026

• CNBC, "Meta debuts Muse Wonder AI model in latest test of users' willingness to pay", 12 Oct 2026