On Monday, October 12, Meta’s Superintelligence Labs shipped Muse Wonder and Muse Wonder Full, the company’s first frontier models, both trained at the new Prometheus data center in central Ohio. The headline for engineers is the Artificial Analysis Intelligence Index score: Wonder Full lands at 56, tying Claude Sonnet 5.5, ahead of GPT-6 Astra, Gemini 4 Argon, and Claude Fable 5.1 (each at 53), and trailing Claude Opus 5.5 at 58. The API-facing Wonder scores 52. Investors noticed; shares climbed as much as 3.4%, still up roughly 3.2% near $766.60 into October 13.

But the detail that reshapes how small businesses get found isn’t on a benchmark table. It’s Active Memory.

Muse Wonder powers the Muse consumer agent, which Meta says will soon roll out to Muse subscribers before any broader release. With Active Memory and Active Focus baked in, Muse can remember what a user asked for last month, a plumber, a caterer, a Pilates studio, and resurface those answers without being re-prompted. “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,” said Kavya Kopparapu, a research scientist at Meta, in a video accompanying the launch. Meta claims Wonder beats Fable, Opus, and Astra on its in-house MemoryBench, a vendor-run benchmark worth reading with the usual skepticism.

Pricing underscores the distribution play. The Meta Model API lists Wonder at $4 per million input tokens, $20 per million output, and $0.20 per million cached inputs, with roughly a one-million-token context window. Only the weaker Wonder is available via API; Wonder Full is gated to Muse app subscribers at $20/month and $100/month, and to Muse Code’s Power Usage ($50/month) and Maximum Usage ($200/month) tiers.

Alexandr Wang, Meta’s AI chief, framed the design philosophy bluntly: “When you do real-world tasks, you’re going to get roadblocks and you’re going to get things that come up or various setbacks. We specifically tried to train the model to be able to try other approaches instead of stalling and then checking back in with the user at any time.”

The structural implication for owners is simple. Muse is becoming a persistent discovery channel, not a one-shot query box. Being the business Muse surfaces the first time compounds into being the business it remembers. That’s a different competitive surface than the small-business agent Meta rolled out earlier this month, and a different pressure than Muse-initiated voice outreach. It’s the 2010 Google Places moment, replayed with memory.

Sources