Skip to content

Meta

Muse Wonder

Meta's most capable models: Wonder, through the Meta Model API, and Wonder Full, for Muse Code subscribers.

Announced
Access
Public

Benchmark scores

TestScoreBest on the trackerHow it was measured
Epoch Capabilities Indexmany tests combined into one score162 pointsGPT-6.2 Astra 174 pointsIndependentlisted as "Muse Wonder", CI 159.6 to 164.67Run 9 Nov 2026
AA Intelligence Index v5about ten tests in one score43.6 pointsClaude Fable 5.2 64.6 pointsIndependentv5.0.1; listed as Muse Wonder (Max)Run 10 Nov 2026
APEX-Agentsprofessional office tasks66%Gemini 4 Carbon 84%Independentpass@1, max effort; listed as "Muse Wonder 1.0"
FrontierMath T1–3hard maths problems82%Claude Fable 5.2 95%Independentmax reasoning effort (muse-wonder-1.0_max)Run 12 Oct 2026
FrontierMath T4research-level maths problemsSaturated71%GPT-6.1 Sol 100%Independentmax reasoning effort (muse-wonder-1.0_max)Run 12 Oct 2026
OTIS Mock AIMEcompetition mathsSaturated99%GPT-5.5 100%Independentmax reasoning effort (muse-wonder-1.0_max)Run 12 Oct 2026

Compare every model on the same tests →

Sources