Meta launches Muse Wonder frontier AI models for coding and agent use

Traders Union
All News
Financial News
Meta Muse Wonder AI launch

Highlights

Meta's Superintelligence Lab launched Muse Wonder and Muse Wonder Full, benchmarked comparably to Fable 5.1 and GPT-6 Astra, targeting complex coding tasks.
Wonder is priced at $4.00 per million input tokens and $20.00 per million output tokens via the Meta Model API, with Wonder Full exclusive to Muse Code subscribers.
Meta claims Muse Wonder scores higher than Fable, Opus, and Astra on MemoryBench, emphasizing active memory features for advanced personal assistant applications.

Model launch, pricing and training details

As reported by TechCrunch, Meta’s Superintelligence Lab released Muse Wonder and Muse Wonder Full on Monday as its first frontier-model effort, positioning the pair above the lighter Muse Spark model in capability and benchmark performance.

Meta says the new models are designed for more complex tasks such as coding, with Wonder and Wonder Full posting benchmark results comparable to Fable 5.1 and GPT-6 Astra in the company’s testing. In its blog post, Meta describes the models as the next step in its scaling strategy and a major advance toward what it calls personal superintelligence.

That positioning is reflected in pricing through the Meta Model API, where Wonder costs $4.00 per million input tokens and $20.00 per million output tokens. Wonder Full is reserved for Muse Code subscribers and is not available through the API, while both models are also set to power the Muse agent.

Meta says the models were trained at its Prometheus data center in central Ohio, which came online earlier this year.

Memory features and competitive AI positioning

Meta places particular emphasis on what it calls active memory, or the ability of the models to retain and synthesize information supplied by users over time. The company presents that feature as central to its broader push into personal assistants that can handle longer-running tasks.

To support that claim, Meta points to the MemoryBench benchmark, which tests continual learning and memory construction frameworks. According to the company’s testing, Muse Wonder scores higher on MemoryBench than Fable, Opus and Astra.

The release is Meta’s first frontier-model launch since Scale AI founder Alexandr Wang restructured the company’s AI efforts last year. Since that overhaul, Meta has introduced the lightweight Muse Spark model but has largely stayed away from the state-of-the-art releases that have defined competition from OpenAI and Anthropic.

Meta’s push to monetize AI through its Muse assistant and enterprise offerings was the focus of our earlier article, highlighting rapid early adoption and the company’s effort to open new revenue streams beyond advertising. We also outlined the scale of Meta’s rising AI capital expenditures and why investors are watching whether those investments translate into durable cash flows and returns.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.