Zhipu's GLM Legal debuts at #1 on Harvey's Legal Agent Benchmark among open-weight models
Zai_org · x · 2026-10-13
Per Vals AI, Zhipu's new legal-specific GLM Legal debuts as the top open-weight model on Harvey's Legal Agent Benchmark with 19.2% accuracy, ahead of Mistral Large 4 (15.8%) and Kimi K3 (12.9%). It also leads LegalBench at 86.23%, narrowly edging Kimi K3's 86.21%.
Related event: Zhipu Launches GLM-Legal, Tops Harvey Legal Agent Benchmark(2 posts)→
More from Models
- Meta introduces Muse Wonder, frontier model for agentic tasks — saln1 · 2026-10-13
- Meta researcher shows Muse Wonder Full's step-change in out-of-distribution performance — ren_hongyu · 2026-10-13
- Meta launches Muse Wonder and Muse Wonder Full, claiming SOTA on agentic tasks — shengjia_zhao · 2026-10-13
- Leaked Muse Wonder benchmarks top SWE-Marathon at 57.5%, beating Claude Opus 5.5 and GPT-6 — saln1 · 2026-10-13
- Dev hands-on: Muse clearly beats Dots on tool calling, personalization feels natural — evijit · 2026-10-13
- Meta's own Muse assistant insists Muse Wonder doesn't exist hours after launch — msg · 2026-10-13