Zhipu GLM-Legal
Zhipu released a legal-domain model built on GLM-5.3-Flash for contract review, evidence analysis, and cross-border compliance work, topping Vals AI's Harvey LAB open-weights leaderboard.
Why it matters
Topped Vals AI's Harvey LAB open-weights leaderboard with a 19.2% full-task pass rate at a reported 1/70th the cost of comparable closed models.
What you could build with it
A boutique law firm or in-house legal team could deploy GLM-Legal behind a private endpoint to pre-screen vendor contracts — flagging termination clauses, payment terms, and handover obligations — so lawyers review a redlined draft instead of a raw 60-page document.
Does it hold up?
Evidence is mixed and worth stating plainly: it leads open-weight legal benchmarks per Vals AI, but the 19.2% absolute pass rate means roughly 4 in 5 real tasks need human rework, and public weight availability is unclear (API-by-application at launch per English-language coverage). Best treated as a first-draft generator, not a sign-off tool.
Built with Zhipu GLM-Legal
- Z.ai GLM-Legal Beats Claude Fable 5 and Kimi K3 on Harvey LABMarkTechPost · Independent breakdown of Vals AI's benchmark numbers, cost per test ($0.15 vs $19.23 for Claude Fable 5), and the current access model.
- 《新东西观察》第33期：法律智能体只过 19.2%，先别当律师智能体世界 · Independent stress-test notes: 19.2% pass rate means most real tasks still need human rework, single tasks take ~37 minutes, and it cannot read scanned images.
- Can You Spot the $0.15 Legal AI? (GLM-Legal Test)youtube · Blind-test video pitting GLM-Legal against its base model and Claude on contract review, evidence analysis, and a privacy plan, with citation fact-checks and the Harvey LAB leaderboard for context.
- GLM-Legal: 19.2% on Harvey LAB, beta by applicationorcarouter · Close read of Z.ai's three launch demos, analyzing where GLM-Legal's contract redlines and evidence arguments diverge from the base GLM-5.3-Flash model.
Learn more
First spotted on article: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73NVIDIA Isaac Lab 3.0 GA + Isaac OS open runtime + Agile ONE SN humanoidMajor Isaac platform expansion at GTC Berlin Oct 21: Isaac Lab 3.0 is generally available as the open-source robot…models · JEV 0.73StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73NVIDIA Nemotron 3.5 Retriever collection: Embed 8B/1B, Rerank 8B, Parse 631MNVIDIA's Oct 21 collection drop for agentic retrieval: Nemotron-3.5-Embed in 8B BF16, 1B BF16 and 1B NVFP4…models · JEV 0.72Kimi K3.1 Pro: Moonshot's 2.8T open-weight flagship officially launchesMoonshot AI officially launched Kimi K3.1 Pro on Oct 23, a mixture-of-experts model with 2.8 trillion total / 104…models · JEV 0.72
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.