Summary
Meta has released Muse Wonder, a family of two models that the company says belongs alongside the best from OpenAI, Anthropic, and Google. This video from Paul J Lipsky explains what the models are, what Meta says they can do, and why the launch matters for the Muse app.
Muse Wonder and the larger Muse Wonder Full were built for agentic work, with examples aimed at areas like law, finance, and health, such as handling an insurance claim or reviewing a lease. Two features stand out. Active memory carries what the model knows about you from one conversation to the next, and active focus is meant to keep agents working through errors instead of giving up, while still checking in before important actions.
The video walks through Meta’s demos: a browser-based running plan built from calendar and map data, a custom birthday shirt for a cat ordered from the cat’s Instagram photo, the Muse Kitchen game, a travel illness question answered from email and calendar context, an insurance claim email, and a wedding seating chart. On benchmarks, Meta reports Muse Wonder Full scoring 59.1 on its personal super intelligence index and 57.5 on SWE Marathon versus 46.3 for Opus 5.5, while Artificial Analysis independently places it at 56, still behind Claude Opus 5.5.
📺 Source: Paul J Lipsky · Published October 12, 2026
🏷️ Format: News Analysis
