Model reference · open weights

glm-4

LLMs zai-org Text gen 1 build Its own licence terms 11k dl/mo

glm-4 is an open-weight language model from zai-org. glm-4-9b (BF16) weighs 18.8 GB; the smallest configuration that runs it is 2× RTX 3060 12 GB.

  • GLM-4-9B is a 9.4B parameter text-generation model developed by zai-org.
  • It supports Chinese and English with an 8K context length for the base version.
  • The model is released under an "other" licence.

Summary of the zai-org/glm-4-9b model card, 2026-10-01

What it is

Released byzai-org
Released2024-06-04
Parameters9.4B
VRAM18.8 GB for the weights

What it runs on

Memory and cards for glm-4-9b (BF16)

18.8 GBweights, file size
41 MBcache per 1K tokens
764 MBruntime overhead, at least
CardRequests at onceContext maxMemory
8K each32K each
RTX 3060 12 GB … RTX 4060 Ti 16 GB
2 smaller cards
———
RTX 3090 24 GB11291K23.4 GB
RTX 4090 24 GB11290K23.4 GB
RTX 5090 32 GB348128K31.0 GB
L40S 48 GB7218128K44.0 GB
A100 80 GB17443128K78.2 GB
H100 80 GB16140128K78.1 GB
RTX PRO 6000 Blackwell 96 GB20852128K93.8 GB
DGX Spark (GB10) 128 GB unified24862128K107 GB
H200 141 GB33984128K138 GB
B200 180 GB451112128K176 GB
2× RTX 3060 12 GB
tensor parallel
8269K11.6 GB a card
2× RTX 4060 Ti 16 GB
tensor parallel
317128K15.4 GB a card
Memory needed at each load
Requests at once8K tokens each32K tokens each
119.9 GB20.9 GB
521.2 GB26.3 GB
822.2 GB30.3 GB
1624.9 GB41.0 GB
3230.3 GB62.5 GB
6441.0 GB105 GB

One card, with vLLM's small-card settings.

From the model card

What zai-org says about glm-4

Read the model card

Read this in English

2024/08/12, 本仓库代码已更新并使用 transformers>=4.44.0, 请及时更新依赖。

GLM-4-9B 是智谱 AI 推出的最新一代预训练模型 GLM-4 系列中的开源版本。 在语义、数学、推理、代码和知识等多方面的数据集测评中, GLM-4-9B 及其人类偏好对齐的版本 GLM-4-9B-Chat 均表现出超越 Llama-3-8B 的卓越性能。除了能进行多轮对话,GLM-4-9B-Chat 还具备网页浏览、代码执行、自定义工具调用(Function Call)和长文本推理(支持最大 128K 上下文)等高级功能。本代模型增加了多语言支持,支持包括日语,韩语,德语在内的 26 种语言。我们还推出了支持 1M 上下文长度(约 200 万中文字符)的 GLM-4-9B-Chat-1M 模型和基于 GLM-4-9B 的多模态模型 GLM-4V-9B。GLM-4V-9B 具备 1120 * 1120 高分辨率下的中英双语多轮对话能力,在中英文综合能力、感知推理、文字识别、图表理解等多方面多模态评测中,GLM-4V-9B 表现出超越 GPT-4-turbo-2024-04-09、Gemini 1.0 Pro、Qwen-VL-Max 和 Claude 3 Opus 的卓越性能。

我们在一些典型任务上对 GLM-4-9B 基座模型进行了评测,结果如下:

ModelMMLUC-EvalGPQAGSM8KMATHHumanEval
Llama-3-8B66.651.2-45.8--
Llama-3-8B-Instruct68.451.334.279.630.062.2
ChatGLM3-6B-Base61.469.0-72.325.7-
GLM-4-9B74.777.134.384.030.470.1

更多推理代码和依赖信息,请访问我们的 github 。

本仓库是 GLM-4-9B 的基座版本,支持8K上下文长度。

协议

GLM-4 模型的权重的使用则需要遵循 LICENSE。

引用

如果你觉得我们的工作有帮助的话,请考虑引用下列论文。

@misc{glm2024chatglm,
      title={ChatGLM: A Family of Large Language Models from GLM-130B to GLM-4 All Tools},
      author={Team GLM and Aohan Zeng and Bin Xu and Bowen Wang and Chenhui Zhang and Da Yin and Diego Rojas and Guanyu Feng and Hanlin Zhao and Hanyu Lai and Hao Yu and Hongning Wang and Jiadai Sun and Jiajie Zhang and Jiale Cheng and Jiayi Gui and Jie Tang and Jing Zhang and Juanzi Li and Lei Zhao and Lindong Wu and Lucen Zhong and Mingdao Liu and Minlie Huang and Peng Zhang and Qinkai Zheng and Rui Lu and Shuaiqi Duan and Shudan Zhang and Shulin Cao and Shuxun Yang and Weng Lam Tam and Wenyi Zhao and Xiao Liu and Xiao Xia and Xiaohan Zhang and Xiaotao Gu and Xin Lv and Xinghan Liu and Xinyi Liu and Xinyue Yang and Xixuan Song and Xunkai Zhang and Yifan An and Yifan Xu and Yilin Niu and Yuantao Yang and Yueyan Li and Yushi Bai and Yuxiao Dong and Zehan Qi and Zhaoyu Wang and Zhen Yang and Zhengxiao Du and Zhenyu Hou and Zihan Wang},
      year={2024},
      eprint={2406.12793},
      archivePrefix={arXiv},
      primaryClass={id='cs.CL' full_name='Computation and Language' is_active=True alt_name='cmp-lg' in_archive='cs' is_general=False description='Covers natural language processing. Roughly includes material in ACM Subject Class I.2.7. Note that work on artificial languages (programming languages, logics, formal systems) that does not explicitly address natural-language issues broadly construed (natural-language processing, computational linguistics, speech, text retrieval, etc.) is not appropriate for this area.'}
}

Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.

© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms