Z.ai’s GLM-Legal scores 19.2% on Harvey LAB, outpacing Claude Fable 5 and Kimi K3
Z.ai released GLM-Legal, a legal-focused AI model built on its GLM-5.3-Flash base, achieving a 19.2% all-pass rate on the Harvey Legal Agent Benchmark (LAB)—nearly triple the 6.7% of its base model. The model targets contract review, evidence analysis, and compliance research, scoring 7 of 79 on Vals AI’s leaderboard, just 2.3 points behind the top-ranked Muse Spark 1.2 at 25.42%. It also…
Key points
- GLM-Legal scores **19.2%** on Harvey LAB, up from **6.7%** for its base model, ranking **#7 of 79** behind 6 Meta/Google models.
- Costs **$0.15 per test** on Harvey LAB, **128x cheaper** than Claude Fable 5’s **$19.23** per equivalent run, per Vals AI.
- API access is **application-only**, with no published pricing, open weights, or self-hosting support yet.
Access is restricted to an application-only API, with no published pricing or open weights. Z.ai claims parity with frontier models on LegalBench (86.23% vs. Claude’s 88.56%), but performance drops on Legal Research Bench (41.83%), below its base model’s 45.19%. The model runs on Z.ai’s hosted API and is not yet available for self-hosting. Developers can prototype using the base model ($0.15/$0.50 per 1M tokens) before swapping to GLM-Legal, though no release date for broader access has been set.
Model pages: Kimi K3 → · GPT-6 Astra →
The story so far
2 episodes →- Z.ai’s GLM-Legal scores 19.2% on Harvey LAB, outpacing Claude Fable 5 and Kimi K3this story
Z.ai GLM-Legal Beats Claude Fable 5 and Kimi K3 on Harvey LAB
MarkTechPost · 13 October 2026
Loading the full article…
This text was published by MarkTechPost and written by Michal Sutter. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Microsoft releases Decision-1 AI model starting at $0.042 per million input tokens · 1 src
- Google launches Gemini 4 Argon model · 14 src
- TypeSafe’s Jev sparks copycat decision models after researcher claims prior work · 1 src
- Cloudflare releases Clef-omni, a 30B multimodal decision model for text, image, audio and video · 1 src
- vLLM Semantic Router releases Decision 3.0 multimodal models with 5 sizes under Apache-2.0 · 1 src
Comments
via GitHub Discussions