GLM-Legal

Z.ai · released Oct 13, 2026

Unknown

Openness taxonomy

access, licence, and disclosure — never a single open/closed binary

Weights availability: unknown
Licence: unknown
Licence category: unknown
Commercial use allowed: unknown
Derivatives allowed: unknown
Redistribution allowed: unknown
Field-of-use restrictions: unknown
Training code availability: unknown
Inference code availability: unknown
Architecture disclosure: unknown
Training data disclosure: unknown

Deployability

raw values as extracted — free text, never inferred from branding

Self-hosting: unknown
On-premise: unknown
Edge deployability: unknown
API availability: Access by application
Regions available: unknown

Benchmarks

as reported — vendor-supplied unless otherwise noted

Harvey Legal Agent Benchmark (Harvey LAB), all-pass 19.2%

National-security pathway & evidence

from the assessed source item, when one is linked

Impact: lowConfidence: lowNat-sec: noneindependent reporting

Evidence limitations: The item is a short single-source report; it provides no benchmark methodology or independent confirmation, and access is described only as by application.

Z.ai GLM-Legal Beats Claude Fable 5 and Kimi K3 on Harvey LAB

Linked research papers

from paper_model_links

LongHarness: Advancing Agentic Reasoning over 100M-Token Corpora via Stateful Evidence Managementevaluates · 100% confidence
Adaptive DHiSparse: Speculative Decoding over a Load-Adaptive KV Hierarchyevaluates · 99% confidence
Harness^2: Recursive Agent Harnessing for an Open Worldevaluates · 100% confidence
EvoCowork: A Benchmark for Real-World Office Tasksevaluates · 100% confidence
DataPraxis: Scaling Analytics Tasks via Verified Workspace Evolutionevaluates · 80% confidence
HyScale: Scaling Long-Context Training for Learned Sparse Attentionevaluates · 90% confidence