Benchgen
Models/zhipu-ai/

GLM-5.1

DraftPublic
Model CardTraining jobsEvaluations

Model Details

GLM-5.1

Quick answer: GLM-5.1 is Zhipu AI's April 2026 MIT model scoring 79.3% BrowseComp, 86.2% GPQA-Diamond, and 68.7% CyberGym. Significant upgrade from GLM-5 — open-source agentic reasoning.

What GLM-5.1 Is

GLM-5.1 is Zhipu AI's April 2026 flagship, switching from Proprietary (GLM-5) to MIT license. The 86.2% GPQA-Diamond and 68.7% CyberGym demonstrate strong scientific and cybersecurity reasoning. BrowseComp 79.3% is competitive with frontier models.

Specifications

FieldValue
OrganizationZhipu AI
LicenseMIT
HuggingFacezai-org/GLM-5.1
Release dateApril 7, 2026
Knowledge cutoffDecember 2025
ModalityText only

Public Benchmark Scores

BenchmarkScoreSourceDate
BrowseComp79.3%Benchgen evaluation2026-04
GPQA-Diamond86.2%Benchgen evaluation2026-04
CyberGym68.7%Benchgen evaluation2026-04

Pricing

Available via Zhipu AI API (bigmodel.cn).

GLM-5.1 vs Alternatives

ModelBrowseCompGPQA-DiamondCyberGymLicense
GLM-5.179.3%86.2%68.7%MIT
GLM-575.9%—43.2%Proprietary

GLM-5.1 vs GLM-5: BrowseComp +3.4% (79.3% vs 75.9%), CyberGym dramatically improved (68.7% vs 43.2%). GLM-5.1 also switches to MIT from Proprietary.

Frequently Asked Questions

What is GLM-5.1? Zhipu AI's April 2026 MIT model scoring 79.3% BrowseComp, 86.2% GPQA-Diamond, 68.7% CyberGym. Significant upgrade from GLM-5.

Scores from Benchgen evaluations. Last updated 2026-07-24.

Benchmark Leaderboards

This model isn’t on any benchmark leaderboard yet.