glm-4-9B-chat-hf: benchmark results
Zhipu's GLM-4-9B-Chat release for the native Transformers GlmForCausalLM implementation, supporting 128K context; distinct from the base and 1M-context releases. Provider: Zhipu. Access: Open.
Aggregate Rating 1468 ± 44, rank #3133 of 10556 rated models, from 16 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Portuguese LLM - ASSIN2 RTE | 92.72 | Macro F1 (%) | 82.5 |
| FaithJudge - RAGTruth Summarization | 9.33 | Hallucination rate (%) on 150 RAGTruth summaries | 79.7 |
| Open LLM Leaderboard - IFEval | 65.13 | Score | 77.5 |
| Open Portuguese LLM - FaQuAD NLI | 75.71 | Macro F1 (%) | 74.3 |
| FaithJudge - FaithBench Summarization | 38.89 | Hallucination rate (%) on 72 FaithBench summaries | 68.9 |
| Open Portuguese LLM - BLUEX | 56.05 | Accuracy (%) | 65.1 |
| Open Portuguese LLM - OAB Exams | 46.24 | Accuracy (%) | 63.3 |
| Open LLM Leaderboard - GPQA | 30.29 | Score | 59.3 |
| Open Portuguese LLM - ENEM | 62.07 | Accuracy (%) | 50.9 |
| Open LLM Leaderboard - MATH Level 5 | 8.46 | Score | 42.4 |
| FaithJudge | 25.44 | Hallucination rate (%), pooled over 511 RAG responses (Faith | 32.4 |
| Open LLM Leaderboard - BBH | 44.32 | Score | 32.4 |
Interactive version: theaggregate.ai/model?slug=glm-4-9b-chat-hf · How It Works · Data refreshed daily, snapshot 2026-11-17.