Legal-domain AI model for contract review, evidence analysis and compliance research, trained for agentic legal work.
Official announcement - Advancing Agentic Legal Intelligence
1 / 3Overview
What is GLM-Legal?
Z.ai's legal AI model for contract review and compliance
GLM-Legal is Z.ai's legal AI model for contract review and compliance, built for agentic legal work. Based on GLM-5.3-Flash, it went through domain mid-training and reinforcement learning across tens of thousands of real legal environments. It extracts termination and handover clauses and proposes actionable negotiation suggestions, analyzes large document sets while linking external statements with internal records and citing sources, and produces privacy compliance checklists for multiple markets. On independent benchmarks it achieved a 19.2% full pass rate on Harvey LAB (12.5 points above its base model) and frontier closed-model level on LegalBench and PLawBench, evaluated by Vals. It is currently in application-based beta API access.
Platforms and languages
GLM-Legal Availability
Platforms
Languages
Capabilities
GLM-Legal Key Features
Legal-domain model
Based on GLM-5.3-Flash, trained with domain mid-training and reinforcement learning across tens of thousands of real legal environments.
Contract review
Extracts termination and handover clauses and proposes actionable negotiation suggestions.
Evidence analysis
Analyzes large document sets, links external statements with internal records, cites sources, and distinguishes facts from inferences.
Compliance research
Privacy compliance checklists for multiple markets, including SG PDPA, MY PDPA and ID PDP Law.
Independent benchmarks
19.2% full pass rate on Harvey LAB (+12.5pp over the base model) and frontier closed-model level on LegalBench/PLawBench, independently evaluated by Vals.
Best for
Who uses GLM-Legal?
Law firms, legal-tech companies, in-house legal teams and solo practitioners
Review contracts, analyze evidence and research compliance with a legal-domain AI model
Common questions
GLM-Legal FAQs
How can I get access?
It is in beta with application-based API access; apply through Z.ai's legal consultation page.
Are the model weights public?
Not announced; it is currently an application-based beta.
How good is it, really?
19.2% full pass rate on Harvey LAB and frontier closed-model level on LegalBench/PLawBench, independently evaluated by Vals.
Where can I use it?
Via the beta API; it is also coming to the AutoClaw desktop agent platform.
Reviews
See what the community thinks and share your experience.
—
Based on 0 ratings
Rating distribution
Community reviews
No reviews
No written reviews yet.
