Meta releases Muse Wonder, its first frontier-class AI model

Meta's new Muse Wonder and Muse Wonder Full, released October 12 by its Meta Superintelligence Labs unit under former Scale AI chief Alexandr Wang, land on the Artificial Analysis Intelligence Index at 52 and 56. That puts them roughly level with Claude Sonnet 5.5 at 56 and GPT-6 Astra at 53, and a point behind Claude Opus 5.5 at 58. Wired reported that Meta's biggest model "rivals the world's best" and ships with new features to keep agents from going off the rails.

The two capabilities Meta is marketing hardest are "Active Memory" and "Active Focus". The first carries context across sessions; the second is meant to keep multi-step agents from drifting as they navigate between apps. "We put a tremendous amount of effort into making sure your model can remember what's important to you over days, weeks, months, and beyond," research scientist Kavya Kopparapu said. Meta's launch post calls the pair "the next step on our scaling ladder and a major leap toward personal superintelligence."

Three of the AI researchers on our radar shared the Wired write-up the day it ran. The commercial twist is that only the smaller model is sold the usual way. According to VentureBeat, Muse Wonder lists at $4 per million input tokens and $20 per million output tokens on the Meta Model API, roughly in line with GPT-6 Sol and Opus 5.5. Muse Wonder Full, the version that scores 79.0 on DeepSWE 1.1 and leads the pair on coding, is restricted to Muse Code subscribers at $50 per month for the Power tier and $200 per month for the Maximum tier.

Both models were trained at Meta's Prometheus data center in Ohio. Meta says Wonder scores higher on MemoryBench than Fable, Opus, or Astra, though the launch materials do not publish the raw per-benchmark numbers behind that claim.

The release follows a roughly 14-month build-out under Wang that Meta has positioned as a bet on frontier capability rather than on open weights.