GLM-4 32B: benchmark results
Zhipu's open GLM-4 32B model (0414 release). Provider: Zhipu. Released 2025-04-14. Access: Open.
Aggregate Rating 1595 ± 15, rank #459 of 1656 rated models, from 53 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| FoodGuardBench - Vanilla Harmful Queries | 10 | Attack success rate (%) of the plain, non-jailbroken harmful | 100 |
| VisualWebArena | 48.7 | Success Rate (%) | 93.5 |
| EComStage - Scenario Route | 87.8 | Task Score (%) | 90.6 |
| EComStage - RAG-QA | 69.87 | Task Score (%) | 87.5 |
| SWE-Arena | 1002 | Elo Score | 81.9 |
| EComStage - Attitude Classification | 83.73 | Task Score (%) | 81.2 |
| EHRBench | 66.12 | Overall Acc (%) (self-reported) | 78.3 |
| LLMEval-Fair - Management | 9.13 | Discipline Score (0-10) | 76.7 |
| LLMEval-Fair - Medicine | 8.23 | Discipline Score (0-10) | 75.9 |
| EComStage | 82.63 | Task Score (%) | 75 |
| LLMEval-Fair - Economics | 8.97 | Discipline Score (0-10) | 69.8 |
| LLMEval-Fair - History | 8.77 | Discipline Score (0-10) | 69 |
Interactive version: theaggregate.ai/model?slug=glm-4-32b · How It Works · Data refreshed daily, snapshot 2026-10-20.