Meta releases its first frontier model, Muse Wonder
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
Meta now fields a frontier-tier model, Muse Wonder, at $4/$20 per million input/output tokens — competitive on benchmarks with GPT-6 Astra and Fable 5.1, and priced in line with GPT-6 Sol and Opus 5.5. The differentiator is "active memory": it tops MemoryBench for cross-session recall, making it worth evaluating for long-running agents that need persistent user context, though the stronger Wonder Full variant is locked to Muse Code subscribers and unavailable via API.
Muse Wonder is priced at $4 per million input tokens and $20 per million output tokens via Meta’s API, putting Meta’s first frontier model in the same cost tier as top commercial models while claiming comparable benchmark performance to Fable 5.1 and GPT-6 Astra. The practical differentiator is active memory: Meta is positioning Wonder for long-running agents that retain and synthesize user context over time, while the more capable Wonder Full is locked to Muse Code subscribers and unavailable through the API.