Muse Wonder Full
The bigger Muse Wonder — Meta's top-scoring model, sold only inside Muse Code plans, never through the API.
About
Muse Wonder Full, released October 12, 2026 alongside the standard Muse Wonder, is the more capable of Meta's two frontier models — and the one you cannot buy with an API key. It is available only inside Muse Code: the $50-a-month Power Usage plan includes it, with a $200-a-month Maximum Usage tier above that.
Most of the Wonder family's benchmark wins in Meta's launch table are Full's: it is the model Meta compared against GPT-6 Astra and Claude Opus 5.5, and on Artificial Analysis's independent index it scores 56 against 52 for the API-available Wonder and 58 for Opus 5.5. Meta trains on your code by default inside Muse Code unless you switch it off, which matters when evaluating the subscription path.
The practical read: the model that wins the headlines is not the model you can build a product on. Standard Wonder is the API model; Full is a Muse Code exclusive.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Independent (Artificial Analysis): 56 vs 52 for standard Wonder, 58 for Opus 5.5. Provider-reported: most of the Wonder family's wins in Meta's 14-row table belong to Full.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Release facts (October 12, 2026; Muse Code-only access; $50/$200 plan tiers; AA index 56) corroborated across Meta's launch materials and independent analyses. Docked for: (1) no API access means no independent production evaluation outside Muse Code; (2) the headline benchmark wins are Meta's own table; (3) proprietary with no weights, and the Muse Code training-on-your-code default is an opt-out, not opt-in.