GLM-4: benchmark results
Zhipu's original GLM-4 API model (dated 0116), pretrained mostly in Chinese and English, the initial release later updated by GLM-4 (0520). Provider: Zhipu. Released 2024-01-16. Access: API.
Aggregate Rating 1539 ± 17, rank #648 of 1656 rated models, from 131 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenCompass Language - Content Critique - Chinese (CompassBench 2405) | 62 | Score (%) | 100 |
| OpenCompass Subjective - Language - Chinese (CompassBench 2405) | 55.6 | Score (%) | 97.4 |
| OpenCompass LLM - Average - English (CompassBench 2402) | 59.7 | Score (%) | 97.1 |
| OpenCompass Language - Content Critique - English (CompassBench 2405) | 56 | Score (%) | 96.1 |
| OpenCompass LLM - Agent - English (CompassBench 2402) | 77.3 | Score (%) | 95.6 |
| OpenCompass Knowledge - Humanities - Chinese (CompassBench 2405) | 56.1 | Score (%) | 94.7 |
| OpenCompass Subjective - Knowledge - Chinese (CompassBench 2405) | 60.9 | Score (%) | 94.7 |
| OpenCompass Agent - CIBench - English (CompassBench 2402) | 73.3 | Score (%) | 94.1 |
| OpenCompass LLM - Agent (CompassBench 2402) | 72.1 | Score (%) | 94.1 |
| OpenCompass LLM - Agent - Chinese (CompassBench 2402) | 66.9 | Score (%) | 94.1 |
| MoralChoice | 100 | MoralChoice Acc (%) | 92.5 |
| OpenCompass Agent - CIBench - English (CompassBench 2405) | 72.2 | Score (%) | 92.1 |
Interactive version: theaggregate.ai/model?slug=glm-4 · How It Works · Data refreshed daily, snapshot 2026-10-20.