AIEvals

Benchmarks

PLawBench

Chinese legal practice: eliciting facts during consultation, structured case analysis and drafting legal documents.

Professional knowledgePLawBench authorsVendor-reported onlyReported by Z.ai

How it is run

The authors provide task-specific expert rubrics for consultation, case analysis and document drafting. Z.ai reports five overall percentages in its GLM-Legal comparison, read 13 October 2026.

What it cannot tell you

Z.ai does not specify the task subset, judge, aggregation, reasoning effort or trial count. Its publisher figures remain separate from the author leaderboard and from LegalBench and Harvey LAB; rubric scores do not establish legal-advice reliability.

Why there is no independent table

No independent suite in this dataset runs PLawBench. The numbers below are the ones model publishers report about their own models, which is a different kind of evidence: each publisher chooses its own harness and effort level, so a row is readable on its own and not against the other rows.

Read this with

Benchmarks in the same category, which usually disagree with this one on purpose.

  • LegalBenchSame category. Legal reasoning tasks assembled from published legal benchmarks, including statutory reasoning and contract analysis.
  • Harvey's Legal Agent BenchmarkSame category. Multi-step legal agent work: retrieval, document review and drafting under a rubric rather than a single answer.
  • Legal Research BenchSame category. Whether an agent can research a real US legal question with case-law search, web search and document retrieval, then produce a supported answer — the multi-source work junior associates and paralegals do.
  • Tax Agent BenchSame category. Whether an agent can research a research-grade US tax question and produce a supported answer — the multi-step process a corporate tax practitioner follows from authority to conclusion.

As reported by each publisher

Kept separate per publisher, because the settings are not shared.

Z.ai

GLM-Legal: legal benchmark evaluation · 12 October 2026

RankModelScore
2GLM-Legal77.3%
4GLM-5.3-Flash73.8%
3Kimi K376.2%
1Claude Fable 577.7%
5GPT-6 Astra69.1%

Ranks compare reported scores in the same published row; ties share a rank.

PLawBench. Chinese legal consultation, case analysis and document drafting; judge, task subset and aggregation are unspecified.

Z.ai’s three-row comparison image, read 13 October 2026. All 15 values retain the published model identities; GLM-5.3-Flash is a comparison column, not GLM 5.3. Harvey LAB is the all-pass rate across 24 practice areas, distinct from criteria pass rate. LegalBench and PLawBench retain the percentages shown in the image; their aggregation, task subsets and judge settings are not specified. Reasoning effort, harness configuration and trial counts are not stated. These are publisher reports, kept separate from independent Vals and Artificial Analysis results. Ranks are derived only within each row, with ties sharing a rank.

z.ai/blog/glm-legal

Sources

Every page used for this benchmark, read on 17 October 2026.