Meta Unveils Muse Wonder AI Models, Claiming They Rival Leading Industry Systems

When Meta launched its Muse personal AI agent in September 2026, it ran on a model called Muse Spark. Six weeks later, the company is already reaching for something more powerful.

On 12 October 2026, Meta announced two new models — Muse Wonder and Muse Wonder Full — describing them as its most capable systems yet for personal and professional agentic tasks. The announcement positions Meta firmly in the race against OpenAI, Anthropic and Google to build AI that doesn’t just answer questions, but actually does things on your behalf.

What Are Muse Wonder and Muse Wonder Full?

The distinction between the two models matters. Muse Wonder is available through the Meta Model API and through Muse Code, Meta’s developer-facing coding product. Muse Wonder Full — the more powerful of the pair — is reserved for Muse Code subscribers, at least for now.

Both models are designed with agentic work in mind. That means multi-step reasoning, tool use, long-horizon planning, and coding — the kind of tasks where an AI doesn’t just respond to a single prompt but carries out a sequence of actions across time. Think less “chatbot” and more “digital assistant that books your appointments, drafts your emails, and writes your code while you’re making a cup of tea.”

Meta Research says the models were built specifically for personal and professional agents. The eventual plan is to integrate them into Muse itself, the consumer-facing agent that Meta launched less than two months ago.

The Numbers Meta Is Putting Forward

Meta has published benchmark results to back up its claims, though it’s worth being clear about what those figures represent. According to Meta’s own testing, as reported by The Verge, Muse Wonder Full scored 56 on the relevant benchmark, compared with 58 for Anthropic’s Claude Opus 5.5 and 53 for several other named competing models. Muse Wonder, the lighter of the two, scored 52.

That puts Muse Wonder Full close to — but not quite at — the top of the leaderboard, at least on the metrics Meta chose to highlight. These comparisons have not been independently verified. The Verge and CNBC both covered the announcement while noting that the competitive picture can shift considerably depending on which benchmark you use and who’s running the tests.

So: competitive, possibly. Definitively state-of-the-art? That’s Meta’s claim, not yet an independent verdict.

What Muse Is Actually Supposed to Do

Muse is not a chatbot. That’s the core of what Meta is trying to communicate. The agent is designed to take actions — planning, handling information, interacting with connected services — rather than simply generating text in response to a question.

It’s a meaningful difference. An AI that can browse, schedule, code, and interact with external tools on a user’s behalf raises a different set of questions than one that sits behind a chat window. Questions about what data it can access, what actions it might take without explicit approval, and who is accountable when something goes wrong.

Meta says Muse is initially available to subscribers. Wider integration of the Muse Wonder models into the consumer product is described as an eventual goal, not an immediate one.

A Crowded and Fast-Moving Field

Meta is not alone in this space. Not even close. OpenAI has been developing agentic capabilities within its own product suite. Anthropic has positioned Claude as a capable tool for complex, multi-step tasks. Google is pursuing similar territory with its Gemini models.

What Meta brings to this contest is scale. Billions of people already use Facebook, Instagram and WhatsApp. If Muse becomes embedded in those products — and Meta has made clear that’s the direction of travel — the potential reach is enormous.

But reach and capability are different things. Developers and businesses evaluating Muse Wonder will want to see independent benchmarks, real-world performance data, and clarity on pricing before committing. As of 12 October 2026, confirmed UK pricing for access to either model has not been published.

Privacy and Accountability Questions

A more capable agent means more access. And more access means more data. Any organisation or individual in the UK using Muse to handle personal information would need to consider their obligations under UK GDPR and the Data Protection Act 2018. That applies whether you’re a sole trader using Muse Code or a larger business thinking about automating customer-facing tasks.

The accountability question is also live. If an AI agent takes an action on a user’s behalf — sends a message, makes a booking, submits a form — and that action turns out to be wrong, the question of who is responsible doesn’t have a tidy answer yet. Users would be wise to check what any agent does on their behalf, rather than assuming it got things right.

What This Means for Kent Residents

No specific launch, trial or public-sector deployment involving Muse Wonder has been announced in Kent. If and when Muse becomes available in the UK more broadly, residents and businesses across the county — from software developers to small traders — could encounter it through Meta’s existing products or developer tools. Any Kent organisation that does adopt the system would need to think carefully about data protection obligations under UK GDPR, especially if the agent is handling personal or sensitive information on their behalf.