Leak: Meta's Fable-class Muse Wonder model rolls out at $4 in / $20 out per Mtok
thesaraharminta · x · 2026-10-13
Multiple credible leaks suggest Meta's long-rumored Fable-class model "Muse Wonder" is rolling out, priced at $4 input / $20 output per million tokens ($0.20 cached input) with 1M context. Screenshots describe it as an agentic model for coding and multi-step work: six reasoning levels, native tool calling, MCP support, long-task planning, sub-agent delegation, and error recovery. A max-reasoning "Muse Wonder Full" variant is exclusive to MuseCode Power and Maximum Usage subscription plans. The model has been in development for months with teasers from Meta staff; not yet officially confirmed.
Related event: Meta unveils Muse Wonder, its strongest model yet for personal agents(76 posts)→
More from Models
- Snap's new decision model handles 384K tokens per call at $0.05/M input, free output — thisdudelikesAI · 2026-10-13
- ThursdAI: OpenAI drops 722 math manuscripts, Haiku 5.5 hits 10 cents, agent leaks bank balance — thursdai_pod · 2026-10-13
- Open weights go big: Reflection AI's 501B Beam, Mistral's 1T Large 4, Aleph Alpha's Kolibri — thursdai_pod · 2026-10-13
- Meta's rumored Muse Wonder beats Claude Opus 5.5 on several agent and coding benchmarks — Dr_Singularity · 2026-10-13
- Wonder Full is a good model, but Meta still trails OpenAI by ~3-6 months — scaling01 · 2026-10-13
- Muse team member celebrates coding work hitting SOTA 'again and again' — jyangballin · 2026-10-13