Muse Wonder claims SOTA on Harvey's legal reasoning benchmark, beating GPT-6 and Claude
shuchaobi · x · 2026-10-14
The Muse team says Muse Wonder is SOTA on legal reasoning: on Harvey's LAB-AA benchmark, Muse Wonder Full (max) scores 13.1% and Muse Wonder (max) 10.0%, ahead of GPT-6 Astra (max) at 8.6%, Claude Fable 5.1 (max) at 6.4%, and Claude Opus 5.5 (max) at 4.2%. Their approach leans on in-context learning: laws vary by country/state and keep changing, so instead of memorizing statutes and cases, the model is trained to retrieve relevant legal material and reason over it in context for accurate, up-to-date results — with long, cumbersome legal documents doubling as a testbed for long-context reasoning. The stated mission is closing the legal literacy gap and leveling the playing field.
More from Models
- Sonnet 5.5 landing on the free tier is more significant than people realize — adonis_singh · 2026-10-14
- Copilot keeps hitting company guardrails on tasks its chat mode handles fine — annetgriffin · 2026-10-14
- Gemini 4 Argon reportedly beats human baseline on all four stealth capability evals — scaling01 · 2026-10-14
- User praises Gemini thought summaries but wants full raw traces — QuixiAI · 2026-10-14
- Gemini 4 Argon lands in OpenCode as team's second most-used model — adonis_singh · 2026-10-14
- Claude Opus 5.5 vs Gemini Argon: Same 3D Island Prompt, Strikingly Different Results — adonis_singh · 2026-10-14