GLM-5: benchmark results
Zhipu's flagship GLM-5 open MoE model (40B active). Provider: Zhipu. Released 2026-02-12. Access: Open.
Aggregate Rating 1733 ± 3, rank #141 of 1656 rated models, from 1318 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AGC-Bench - grapheval_ai_researcher | 1.88 | Dataset z-score | 100 |
| AGC-Bench - schnovel | 2.04 | Dataset z-score | 100 |
| Are LLMs Ready to Assist Physicians? PhysAssis | 69.4 | en_mrs (self-reported) | 100 |
| ArmBench-LLM - Classification | 64.57 | Accuracy (%) | 100 |
| AutoMedBench - Visual Question Answering | 64 | Overall run score (0-100) on the visual question answering t | 100 |
| BALSAM - Information Extraction | 53.79 | Overall score (0-100, LLM-judged generation and multiple cho | 100 |
| BALSAM - Logic | 42.41 | Overall score (0-100, LLM-judged generation and multiple cho | 100 |
| BALSAM - Translation/Transliteration | 70.18 | Overall score (0-100, LLM-judged generation and multiple cho | 100 |
| ClawBench (OpenClaw) - Work | 38.1 | Success rate (%) on the 21 work tasks of the 153-task ClawBe | 100 |
| ClawsBench - Capability | 76.48 | Mean native task score x100 (-100 to 100); OpenClaw, domain | 100 |
| ClawsBench - Safety (Claude Code; skills off; meta off) | 7.54 | Mean native task score x100 (-100 to 100); Claude Code, doma | 100 |
| CodeRouterBench - LLM Router (3-Shot) | 46.03 | Average task score (%) of the backend the model routes each | 100 |
Interactive version: theaggregate.ai/model?slug=glm-5 · How It Works · Data refreshed daily, snapshot 2026-10-20.