# Z.ai GLM-Legal tops Harvey LAB benchmark

> Z.ai’s legal AI model scores 19.2% all‑pass on Harvey’s benchmark, beating Claude Fable 5 and Kimi K3.

Oossa · 2026-10-13 · https://oossa.com/en/z-ai-glm-legal-tops-harvey-lab-benchmark

Z.ai released a legal‑focused AI model called GLM‑Legal. Built on the GLM‑5.3‑Flash architecture, it achieved a 19.2% all‑pass rate on Harvey’s Legal Agent Benchmark. That score is almost three times higher than the model’s baseline. In the test, it outperformed Claude Fable 5 and Kimi K3. The benchmark costs $0.15 per test and access requires an application.

## The facts

- 19.2% all‑pass score on Harvey LAB
- Model built on GLM‑5.3‑Flash; benchmark price $0.15 per test

## Why it matters

Law firms that can apply for access may evaluate a cheaper, higher‑perform AI tool for legal research and document review.

## Sources & references

1. [Z.ai GLM-Legal Beats Claude Fable 5 and Kimi K3 on Harvey LAB](https://www.marktechpost.com/2026/10/13/zai-glm-legal-harvey-lab-legal-ai-model/) – MarkTechPost, 2026-10-13

Last updated: 2026-10-13