Meta has introduced two artificial intelligence models, Muse Wonder and Muse Wonder Full, designed to handle complex tasks involving coding, reasoning, and long-term memory. Announced on October 12, 2026, the models are the company’s first frontier AI releases, with Meta claiming their performance is comparable to leading models from OpenAI and Anthropic.
According to Meta’s official announcement, the Wonder models are larger and more capable than Muse Spark, which was designed for lightweight tasks requiring faster responses. The company says the new models deliver higher benchmark scores and improved performance on complex programming and multi-step assignments.
Muse Wonder Models Focus on Memory and Complex Tasks
Meta’s published model card reports that Muse Wonder and Wonder Full achieve benchmark results comparable to Fable 5.1 and GPT-6 Astra. The models support tasks involving software development, document analysis, and specialized knowledge in fields such as law, finance, and healthcare.
A major feature is Active Memory, which allows the models to retain and retrieve relevant information across conversations over extended periods. Meta designed the system to remember user preferences and important details over days, weeks, or months.
The company evaluated this capability using MemoryBench, a benchmark that measures how well AI systems retain and use information over time. According to Meta’s internal testing, Muse Wonder outperformed Fable, Opus, and Astra on memory-related tasks.
The models also feature Active Focus, which allows AI agents to continue working through difficult assignments, try alternative approaches, and recover from errors. Meta says these capabilities support longer software development projects and personal assistance tasks.
Pricing and Availability of Meta’s New AI Models
Muse Wonder is available through the Meta Model API at $4 per million input tokens and $20 per million output tokens. Its pricing is comparable to GPT-6 Sol and Opus 5.5, placing it among higher-cost commercial AI models.
Muse Wonder Full, the more capable version, is available exclusively to Muse Code subscribers and will not be offered through the API. Meta also plans to integrate the models into its Muse AI agent, with access initially rolling out to subscribers.
The release follows a restructuring of Meta’s AI operations led by Scale AI founder Alexandr Wang in August 2025. Since then, Meta has introduced its lightweight Muse Spark model and Muse agent before releasing the more computationally demanding Wonder models.
Both models were trained at Meta’s Prometheus data center in central Ohio, which began operating earlier in 2026. The facility provides the computing infrastructure used to develop Meta’s larger AI systems.
Featured image credits: Wikimedia Commons
For more stories like it, click the +Follow button at the top of this page to follow us.