glm-4-9B: benchmark results
Zhipu AI's GLM-4-9B, the base pretrained checkpoint in the GLM-4 series, supporting an 8K context window; a separate GLM-4-9B-Chat instruction-tuned version also exists. Provider: Zhipu. Released 2024-06-05. Access: Open.
Aggregate Rating 1470 ± 13, rank #889 of 1656 rated models, from 105 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SalesLLM - Chinese (CustomerLM User) | 7.14 | SalesLLM benchmark score (1-10): 0.5 x buying-intent score ( | 100 |
| Open PL LLM - PSC (generative, 5-shot) | 98.94 | Binary F1 (%) | 99.4 |
| Open Portuguese LLM - FaQuAD NLI | 83.14 | Macro F1 (%) | 96.2 |
| Open PL LLM - PPC (multiple choice, 5-shot) | 79.8 | Accuracy (%) | 92.3 |
| Open PL LLM - PoQuAD Open Book (generative, 5-shot) | 71.22 | Levenshtein Similarity (%) | 90.8 |
| Open Portuguese LLM - ASSIN2 RTE | 93.34 | Macro F1 (%) | 88.8 |
| Open Portuguese LLM - OAB Exams | 55.95 | Accuracy (%) | 88.6 |
| Open Portuguese LLM - BLUEX | 64.53 | Accuracy (%) | 85 |
| Open Portuguese LLM - ENEM | 73.27 | Accuracy (%) | 84.7 |
| Open PL LLM - PolQA Reranking (multiple choice, 5-shot) | 80.61 | Accuracy (%) | 84.3 |
| PilotBench - Instruction Following | 96.45 | Instruction-following score (%): 0.5 x field completeness pl | 83.3 |
| Open PL LLM - RAG | 71.01 | Average RAG Score (%) | 80.4 |
Interactive version: theaggregate.ai/model?slug=glm-4-9b · How It Works · Data refreshed daily, snapshot 2026-10-20.