GLM-4 (0520): benchmark results
Zhipu's GLM-4 (0520), an incremental update to their GLM-4 API model, pretrained mostly in Chinese and English and aligned predominantly for Chinese use. Provider: Zhipu. Released 2024-06-05. Access: API.
Aggregate Rating 1552 ± 25, rank #612 of 1635 rated models, from 92 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SuperCLUE General (June 2024) - Hard | 79 | Score | 92.2 |
| OpenCompass Knowledge - Social Science - Chinese (CompassBench 2407) | 80 | Score (%) | 88.2 |
| SuperCLUE General (June 2024) - Overall | 76 | Score | 87.5 |
| OpenCompass Knowledge - Social Science - English (CompassBench 2407) | 87.2 | Score (%) | 85.3 |
| OpenCompass LLM - Code (CompassBench 2407) | 56.3 | Score (%) | 85.3 |
| OpenCompass LLM - Code - Chinese (CompassBench 2407) | 55.8 | Score (%) | 85.3 |
| OpenCompass Reasoning - Science and Engineering - English (CompassBench 2407) | 51 | Score (%) | 83.8 |
| OpenCompass Code - LCBench - English (CompassBench 2407) | 52.4 | Score (%) | 82.4 |
| OpenCompass Reasoning - Humanities - English (CompassBench 2407) | 49.9 | Score (%) | 82.4 |
| OpenCompass Code - HumanEval - Chinese (CompassBench 2407) | 76.2 | Score (%) | 80.9 |
| OpenCompass Code - LCBench - Chinese (CompassBench 2407) | 43.6 | Score (%) | 79.4 |
| SuperCLUE General (August 2024) - Hard | 65.28 | Score | 78 |
Interactive version: theaggregate.ai/model?slug=glm-4-0520 · How It Works · Data refreshed daily, snapshot 2026-11-05.