Meta releases its first frontier model, Muse Wonder

Meta’s Superintelligence Lab unveiled two new frontier models on Monday—Muse Wonder and Muse Wonder Full—marking the company’s first foray into large‑scale, high‑performance AI systems. Unlike the earlier lightweight Muse Spark, the Wonder pair is built for more demanding tasks, delivering higher benchmark scores and superior results on complex applications such as coding. In Meta’s own testing, both models achieved performance on par with leading competitors like Fable 5.1 and GPT‑6 Astra, and they are accessible via the Meta Model API at $4 per million input tokens and $20 per million output tokens, a price point comparable to GPT‑6 Sol or Opus 5.5. The launch also coincided with a promotional event on October 13, offering discounted passes for attendees.

A central focus of the Wonder models is “active memory,” a capability designed to retain and synthesize user‑provided information over extended periods. Meta highlighted its performance on the MemoryBench benchmark, which evaluates continual learning and memory construction, noting that Muse Wonder outscored rivals such as Fable, Opus, and Astra. Research scientist Kavya Kopparapu emphasized the effort put into ensuring the model can remember important details for days, weeks, or even months, positioning the technology as a step toward personal superintelligence. The models were trained at Meta’s new Prometheus data center in central Ohio, a facility that became operational earlier this year.

The release follows a restructuring of Meta’s AI strategy in August, when Scale AI founder Alexandr Wang reorganized the company’s efforts, leading to the earlier launch of Muse Spark and the Muse agent. By introducing a frontier model, Meta joins the ranks of OpenAI and Anthropic, which have regularly announced state‑of‑the‑art systems. The new models are expected to underpin Meta’s ambitions for long‑term personal assistants and could influence the broader AI market by offering a competitively priced, memory‑enhanced alternative to existing high‑end offerings.

⚡ Effects Interpreter