Meta's Muse Wonder Full beats Opus 5.5 on agent benchmarks: 57.5% on SWE-Marathon
kimmonismus · x · 2026-10-13
Meta claims Muse Wonder Full outperforms leading OpenAI and Anthropic models on sustained agent work: 79.0% on DeepSWE 1.1 and 57.5% on SWE-Marathon vs 46.3% for Claude Opus 5.5 and 42.5% for GPT-6 Astra. It also tops the Personal Superintelligence Index (59.1), AutomationBench-AA (45.4) and MemoryBench (85.6). New mechanisms Active Memory (cross-session context retrieval) and Active Focus (working through blockers) are credited. Wonder is available via Meta's API and Muse Code; Full for Muse Code subscribers. On the Artificial Analysis Index it scores 56, second only to Opus 5.5 (58).
Related event: Meta Unveils Muse Wonder, Its Most Capable Frontier Model(92 posts)→
More from Models
- ReplaySSM ships in GLM-5.3-Flash and Kimi-K3, speeding decoding up to 2.1x — SonglinYang4 · 2026-10-13
- AI2's TutorMoments eval: GPT-6 models over-help least, detailed prompts beat "tutor well" — allen_ai · 2026-10-13
- AI2's TutorMoments benchmarks AI tutors on 520 real moments; human tutors score 0.458 — allen_ai · 2026-10-13
- Allen AI's TutorMoments Benchmark Finds LLM Tutors Over-Help Students — allen_ai · 2026-10-13
- DeepSeek's flash endpoint may be quietly testing a new model, users speculate v4.1-pro — teortaxesTex · 2026-10-13
- 3D game demos hint at AI creativity that transfers to CAD, docs and UI tasks — Angaisb_ · 2026-10-13