GLM-Legal
Z.ai's legal agent model built on GLM-5.3-Flash, scoring 19.2% all-pass on Harvey's LAB benchmark.
This entry is marked for review. Check the original sources before relying on its details.
Overview
Z.ai announced GLM-Legal on October 12, 2026, a legal-domain model built on the GLM-5.3-Flash base (reported as a 320B mixture-of-experts with 18B active parameters and a 1M-token context). Z.ai reports a 19.2% all-pass rate on Harvey's LAB, which covers law-firm workflows across 24 practice areas and counts a task only if every step is correct. The figure is Z.ai-reported and not independently replicated. There is no public API yet; access is via an application form and the AutoClaw desktop agent.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Key features
- 19.2% all-pass on Harvey LAB (Z.ai-reported), a 12.5-point gain over GLM-5.3-Flash
- Built with legal-specific mid-training and reinforcement learning in legal environments
- 1M-token context window (per secondary reporting)
- Available in AutoClaw, Z.ai's desktop agent
Use cases
- Contract and issue-spotting workflows for law teams
- Legal research and drafting agents
- Chinese-law tasks (Z.ai cites PLawBench)
Pricing
No public API pricing; API access by application. AutoClaw signup credits of $24 reported.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- 12 Oct 2026
- Record updated
- 15 Oct 2026
- Record last checked
- 15 Oct 2026