Meta’s Muse Wonder Full Pushes Into the AI Frontier Tier

Meta has released Muse Wonder in two variants: Muse Wonder and Muse Wonder Full. According to The Decoder, Muse Wonder Full scores 56 points on the Artificial Analysis Intelligence Index v4.3, while Muse Wonder scores 52, up from 48 for Muse Spark 1.3. Muse Wonder Full places in the top three on six out of ten benchmarks, signaling a notable jump from Meta’s previous mid-tier position.

The strongest improvements show up in agentic evaluations, which test how well a model can handle complex tasks independently. Terminal-Bench 4.0 performance rose from 33 to 57 percent, while Muse Wonder Full reaches 68 points on the Coding Agents Index. The model also solved 94 out of 100 problems on ProofBench, up from 58 for its predecessor.

Meta highlights two capabilities: Active Memory and Active Focus. Active Memory is designed to remember key information across conversations, while Active Focus helps the model stay with difficult tasks, try alternate approaches, and ask follow-up questions when needed. Both features support Meta’s broader pitch for a personal AI agent that can work with calendars, messages, and documents.

The performance jump brings a major price increase. Artificial Analysis puts Muse Wonder at $4.52 per task on the Intelligence Index, nearly three times Muse Spark 1.3’s $1.60. Token pricing also rises to $4 per million input tokens and $20 per million output tokens, compared with $1.25 and $4.25 for the predecessor; no token pricing is available yet for Muse Wonder Full.