Z.ai's GLM-Legal Ranks 7th on Harvey Benchmark: Full Review & Pricing (October 2026)

Z.ai released GLM-Legal on October 12, 2026, an open-weight legal agent model that ranks seventh on Harvey's Legal Agent Benchmark (LAB) with an independently confirmed 19.17 percent all-pass rate by Vals AI. Built on the lightweight GLM-5.3-Flash foundation, this domain-specialized model offers a one-million-token context window and competitive pricing for agentic legal work. For broader context, explore our AI Tools by Platform.

GLM-Legal's Benchmark Performance

Z.ai reported that GLM-Legal achieved a 19.2 percent all-pass rate on Harvey's Legal Agent Benchmark (LAB). This performance was independently verified by Vals AI, which confirmed an all-pass rate of 19.17 percent. This places GLM-Legal as the seventh-highest-ranking model out of 79 evaluated on the benchmark, indicating its capability in handling complex legal assignments.

The model's foundation is the lightweight GLM-5.3-Flash architecture. Its training involved an expanded legal mid-training corpus and reinforcement learning across tens of thousands of simulated legal task environments, contributing to its specialized performance.

Key Features and Technical Specifications

GLM-Legal is distinguished by its technical specifications tailored for demanding legal workflows. It features a one-million-token context window, enabling it to process extensive legal documents and complex case information. The model also supports a maximum output of 128,000 tokens, facilitating detailed responses and comprehensive document generation.

A significant aspect of GLM-Legal is its open-weight status, which allows developers and researchers to access and customize the model's underlying architecture. This approach can foster innovation and adaptation within the legal AI community.

Pricing and Availability

Vals AI has listed the pricing for GLM-Legal. Users can expect to pay $0.07 per million input tokens and $0.25 per million output tokens. This structure aims to provide a cost-effective solution for legal firms and individual practitioners requiring advanced AI capabilities.

The model became available on October 12, 2026. Its release generated considerable interest, becoming the day's largest AI discussion on Hacker News by midday UTC, accumulating 563 upvotes.

Demonstrated Legal Workflows

GLM-Legal is designed to support a range of agentic legal tasks. Z.ai has demonstrated its application in several key workflows:

• Contract Review: Automating the analysis of legal contracts to identify clauses, obligations, and potential risks.
• Evidence Contradiction Mapping: Assisting in litigation by identifying inconsistencies and contradictions within large sets of evidence.
• Multi-Jurisdiction Privacy Compliance Research: Streamlining the process of researching and ensuring compliance with privacy regulations across various legal jurisdictions.

These applications highlight the model's potential to enhance efficiency and accuracy in specialized legal operations.

Conclusion

Z.ai's GLM-Legal represents a notable advancement in specialized AI for the legal sector. Its strong performance on the Harvey Legal Agent Benchmark, coupled with its open-weight nature and practical features, positions it as a relevant tool for legal professionals. The model's capabilities in areas like contract review and compliance research suggest its potential to streamline complex legal tasks. Interested parties should consider its technical specifications and pricing for integration into their workflows.

Sources

• GitHub - joylarkin/AI-Coding-Landscape: AI coding models, agents, CLIs, IDEs, AI app builders, open source tooling, benchmarks · GitHub
• AI-Coding-Landscape/aicodingmodels.csv at main · joylarkin/AI-Coding-Landscape · GitHub
• Open-weight AI models are catching up to the frontier. The safety gap remains. | TechCrunch