Meta launches Muse Wonder and Muse Wonder Full, claiming SOTA on agentic tasks
shengjia_zhao · x · 2026-10-13
Meta's Shengjia Zhao announced Muse Wonder and Muse Wonder Full, its best models yet, built for long-running multi-step agentic work. Muse Wonder is available today in the Meta Model API and Muse Code; Muse Wonder Full ships in Muse Code.
Benchmarks shared by Meta show Muse Wonder Full leading Claude Fable 5.1, GPT-6 Astra and Claude Opus 5.5 on Personal Superintelligence Index (59.1), Harvey LAB-AA (13.1), AutomationBench-AA (45.4), DeepSWE 1.1 (79.0) and SWE-Marathon (57.5, ultrahigh), while trailing rivals on HealthBench Pro, OSWorld, Terminal-Bench and the Artificial Analysis Index.
Related event: Meta Unveils Muse Wonder, Its Most Capable Agentic Frontier Model(80 posts)→
More from coding & agent
- Cursor Composer predictions hit ~60% acceptance rate, users back to tab-tab-tab flow — rudrank · 2026-10-13
- Nace AI's Drex 2.0: a 35B decision-only model with 3B active params beats Jev on Decision Index — techNmak · 2026-10-13
- Sentient OS launches on-device AI that learns your life overnight and drives your cursor — Scobleizer · 2026-10-13
- Generator-critic pre-push hook: a second AI model must sign off before code lands — curious_vii · 2026-10-13
- Tracking AI coding costs down to each PR: zeeg's slop dashboard experiment — zeeg · 2026-10-13
- Restless HQ's Self-Improving APIs Let Agents Debug Failures via Live Logs — cneuralnetwork · 2026-10-13