GLM-5 README

The project's README from github.com/zai-org/GLM-5. As one markdown file: README.md.

GLM-5.3 & GLM-5.2 & GLM-5.1 & GLM-5

👋 Join our Wechat or Discord community.
📖 Check out the GLM-5.3-Flash blog, GLM-5.3 blog and GLM-5 Technical report.
📍 Use GLM-5.3 & GLM-5.3-Flash API services on Z.ai API Platform.
🔜 Try GLM-5.3 & GLM-5.3-Flash at z.ai.

Introduction

GLM-5.3 & GLM-5.3-Flash

GLM-5.3 uses the same base model as GLM-5.2 — every gain comes from post-training. Compared with GLM-5.2, it is much better at complex coding and long-horizon tasks:

bench_53_2

GLM-5.3-Flash starts from a newly trained base model, with its architecture and training recipe redesigned around capability and efficiency. For the first time in the GLM series, we introduce a hybrid architecture combining sparse and linear attention, sharply reducing long-context serving costs while preserving precise long-context capabilities. The model also adopts Manifold-Constrained Hyper-Connections (mHC) to further improve scaling efficiency. Together with our latest 30T-token multimodal pre-training corpus, these changes enable GLM-5.3-Flash to deliver more intelligence with less compute.

bench_53

GLM-5.2

GLM-5.2, our latest flagship model for long-horizon tasks. It marks a substantial leap in long-horizon task capability over its predecessor GLM-5.1 and, for the first time, delivers that capability on a solid 1M-token context.

GLM-5.2's new capabilities include:

bench_52

On standard coding benchmarks, GLM-5.2 is the strongest open-source model, improving on GLM-5.1 by a wide margin: 81.0 vs. 62.0 on Terminal-Bench 2.1 and 62.1 vs. 58.4 on SWE-bench Pro. It also closes much of the gap to the closed-source frontier — on Terminal-Bench 2.1 (81.0) it lands within a few points of Claude Opus 4.8 (85.0) — while staying ahead of Gemini 3.1 Pro.

For more detail, check our blog.

GLM-5.1

GLM-5.1 is our next-generation flagship model for agentic engineering, with significantly stronger coding capabilities than its predecessor. It achieves state-of-the-art performance on SWE-Bench Pro and leads GLM-5 by a wide margin on NL2Repo (repo generation) and Terminal-Bench 2.0 (real-world terminal tasks).

bench_51

But the most meaningful leap goes beyond first-pass performance. Previous models—including GLM-5—tend to exhaust their repertoire early: they apply familiar techniques for quick initial gains, then plateau. Giving them more time doesn't help.

GLM-5.1, by contrast, is built to stay effective on agentic tasks over much longer horizons. We've found that the model handles ambiguous problems with better judgment and stays productive over longer sessions. It breaks complex problems down, runs experiments, reads results, and identifies blockers with real precision. By revisiting its reasoning and revising its strategy through repeated iteration, GLM-5.1 sustains optimization over hundreds of rounds and thousands of tool calls. The longer it runs, the better the result.

GLM-5

We are launching GLM-5, targeting complex systems engineering and long-horizon agentic tasks. Scaling is still one of the most important ways to improve the intelligence efficiency of Artificial General Intelligence (AGI). Compared to GLM-4.5, GLM-5 scales from 355B parameters (32B active) to 744B parameters (40B active), and increases pre-training data from 23T to 28.5T tokens. GLM-5 also integrates DeepSeek Sparse Attention (DSA), largely reducing deployment cost while preserving long-context capacity.

Reinforcement learning aims to bridge the gap between competence and excellence in pre-trained models. However, deploying it at scale for LLMs is a challenge due to the RL training inefficiency. To this end, we developed slime, a novel asynchronous RL infrastructure that substantially improves training throughput and efficiency, enabling more fine-grained post-training iterations. With advances in both pre-training and post-training, GLM-5 delivers significant improvement compared to GLM-4.7 across a wide range of academic benchmarks and achieves best-in-class performance among all open-source models in the world on reasoning, coding, and agentic tasks, closing the gap with frontier models.

bench

GLM-5 is purpose-built for complex systems engineering and long-horizon agentic tasks. On our internal evaluation suite CC-Bench-V2, GLM-5 significantly outperforms GLM-4.7 across frontend, backend, and long-horizon tasks, narrowing the gap to Claude Opus 4.5.

realworld_bench

On Vending Bench 2, a benchmark that measures long-term operational capability, GLM-5 ranks #1 among open-source models. Vending Bench 2 requires the model to run a simulated vending machine business over a one-year horizon; GLM-5 finishes with a final account balance of $4,432, approaching Claude Opus 4.5 and demonstrating strong long-term planning and resource management.

vending_bench

Download Model

Model Download Links Model Size Precision
GLM-5.3 🤗 Hugging Face
🤖 ModelScope		 744B-A40B FP8
GLM-5.3-BF16 🤗 Hugging Face
🤖 ModelScope		 744B-A40B BF16
GLM-5.3-Flash 🤗 Hugging Face
🤖 ModelScope		 320B-A18B FP8
GLM-5.3-Flash-BF16 🤗 Hugging Face
🤖 ModelScope		 320B-A18B BF16
GLM-5.2 🤗 Hugging Face
🤖 ModelScope		 744B-A40B BF16
GLM-5.2-FP8 🤗 Hugging Face
🤖 ModelScope		 744B-A40B FP8
GLM-5.1 🤗 Hugging Face
🤖 ModelScope		 744B-A40B BF16
GLM-5.1-FP8 🤗 Hugging Face
🤖 ModelScope		 744B-A40B FP8
GLM-5 🤗 Hugging Face
🤖 ModelScope		 744B-A40B BF16
GLM-5-FP8 🤗 Hugging Face
🤖 ModelScope		 744B-A40B FP8

Serve GLM-5 Series Models Locally

GLM-5.3-Flash

GLM-5.3 and Earlier GLM-5 Models

Note

Fine-tuning GLM-5 Series Models

The GLM-5 series supports fine-tuning with the following frameworks. Feel free to try them out:

Citation

If you find GLM-5 series model useful in your research, please cite our technical report:

@misc{glm5team2026glm5vibecodingagentic,
      title={GLM-5: from Vibe Coding to Agentic Engineering},
      author={GLM-5-Team and : and Aohan Zeng and Xin Lv and Zhenyu Hou and Zhengxiao Du and Qinkai Zheng and Bin Chen and Da Yin and Chendi Ge and Chenghua Huang and Chengxing Xie and Chenzheng Zhu and Congfeng Yin and Cunxiang Wang and Gengzheng Pan and Hao Zeng and Haoke Zhang and Haoran Wang and Huilong Chen and Jiajie Zhang and Jian Jiao and Jiaqi Guo and Jingsen Wang and Jingzhao Du and Jinzhu Wu and Kedong Wang and Lei Li and Lin Fan and Lucen Zhong and Mingdao Liu and Mingming Zhao and Pengfan Du and Qian Dong and Rui Lu and Shuang-Li and Shulin Cao and Song Liu and Ting Jiang and Xiaodong Chen and Xiaohan Zhang and Xuancheng Huang and Xuezhen Dong and Yabo Xu and Yao Wei and Yifan An and Yilin Niu and Yitong Zhu and Yuanhao Wen and Yukuo Cen and Yushi Bai and Zhongpei Qiao and Zihan Wang and Zikang Wang and Zilin Zhu and Ziqiang Liu and Zixuan Li and Bojie Wang and Bosi Wen and Can Huang and Changpeng Cai and Chao Yu and Chen Li and Chengwei Hu and Chenhui Zhang and Dan Zhang and Daoyan Lin and Dayong Yang and Di Wang and Ding Ai and Erle Zhu and Fangzhou Yi and Feiyu Chen and Guohong Wen and Hailong Sun and Haisha Zhao and Haiyi Hu and Hanchen Zhang and Hanrui Liu and Hanyu Zhang and Hao Peng and Hao Tai and Haobo Zhang and He Liu and Hongwei Wang and Hongxi Yan and Hongyu Ge and Huan Liu and Huanpeng Chu and Jia'ni Zhao and Jiachen Wang and Jiajing Zhao and Jiamin Ren and Jiapeng Wang and Jiaxin Zhang and Jiayi Gui and Jiayue Zhao and Jijie Li and Jing An and Jing Li and Jingwei Yuan and Jinhua Du and Jinxin Liu and Junkai Zhi and Junwen Duan and Kaiyue Zhou and Kangjian Wei and Ke Wang and Keyun Luo and Laiqiang Zhang and Leigang Sha and Liang Xu and Lindong Wu and Lintao Ding and Lu Chen and Minghao Li and Nianyi Lin and Pan Ta and Qiang Zou and Rongjun Song and Ruiqi Yang and Shangqing Tu and Shangtong Yang and Shaoxiang Wu and Shengyan Zhang and Shijie Li and Shuang Li and Shuyi Fan and Wei Qin and Wei Tian and Weining Zhang and Wenbo Yu and Wenjie Liang and Xiang Kuang and Xiangmeng Cheng and Xiangyang Li and Xiaoquan Yan and Xiaowei Hu and Xiaoying Ling and Xing Fan and Xingye Xia and Xinyuan Zhang and Xinze Zhang and Xirui Pan and Xu Zou and Xunkai Zhang and Yadi Liu and Yandong Wu and Yanfu Li and Yidong Wang and Yifan Zhu and Yijun Tan and Yilin Zhou and Yiming Pan and Ying Zhang and Yinpei Su and Yipeng Geng and Yong Yan and Yonglin Tan and Yuean Bi and Yuhan Shen and Yuhao Yang and Yujiang Li and Yunan Liu and Yunqing Wang and Yuntao Li and Yurong Wu and Yutao Zhang and Yuxi Duan and Yuxuan Zhang and Zezhen Liu and Zhengtao Jiang and Zhenhe Yan and Zheyu Zhang and Zhixiang Wei and Zhuo Chen and Zhuoer Feng and Zijun Yao and Ziwei Chai and Ziyuan Wang and Zuzhou Zhang and Bin Xu and Minlie Huang and Hongning Wang and Juanzi Li and Yuxiao Dong and Jie Tang},
      year={2026},
      eprint={2602.15763},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2602.15763},
}

GLM-5 (zai-org/GLM-5)

Searching Database…

7.4K ·GitHub
GLM-5
Что это такое?GLM-5 — семейство больших языковых моделей с открытыми весами от Z.ai, созданных для агентной разработки и сложных долгих задач кодинга, масштабирующихся до 744 млрд параметров с разреженным вниманием для эффективного обслуживания длинного контекста.
Это готовое приложение или сырая модель, которую нужно развернуть самому?Это не приложение — оно поставляется в виде загружаемых чекпоинтов с открытыми весами (через Hugging Face или ModelScope), которые вы сами разворачиваете с движком инференса, либо доступ происходит через хостинговый API.
Как запустить GLM-5 на своём железе?Скачайте веса с Hugging Face или ModelScope, затем разверните их с помощью одного из перечисленных движков — SGLang, vLLM, Transformers, KTransformers или Unsloth, — каждый из которых снабжён ссылкой на свой cookbook или руководство в README.
Какое железо нужно GLM-5 для работы?Флагманские модели огромны — до 744 млрд параметров всего, из них 40 млрд активных, — поэтому для самостоятельного хостинга нужна серьёзная мульти-GPU-инфраструктура (или Ascend NPU); меньшая GLM-5.3-Flash (320B-A18B) — более лёгкая альтернатива.
В чём подвох с настройками уровня рассуждений?Управление уровнем рассуждений различается между версиями: GLM-5.2 принимает только high или max, а GLM-5.3 и GLM-5.3-Flash добавляют вариант low и молча используют max по умолчанию, если вы явно не передали low или high, — это может искажать сравнения бенчмарков, если быть невнимательным.
Чем различаются GLM-5, GLM-5.1, GLM-5.2, GLM-5.3 и GLM-5.3-Flash?Каждый выпуск нацелен на более длинные горизонты и лучший кодинг: GLM-5 представил базовую модель на 744 млрд параметров с разреженным вниманием, GLM-5.1 улучшил настойчивость на длинных горизонтах, GLM-5.2 добавил стабильный контекст на 1 млн токенов и внимание IndexShare, а GLM-5.3 (плюс меньшая GLM-5.3-Flash с гибридным вниманием) ещё поднял результаты бенчмарков кодинга и кибербезопасности за счёт пост-обучения.
Насколько хорош GLM-5 по сравнению с закрытыми моделями вроде Claude или Gemini?На Terminal-Bench 2.1 GLM-5.2 набирает 81,0 против 62,0 у GLM-5.1, что в пределах нескольких очков от Claude Opus 4.8 (85,0) и впереди Gemini 3.1 Pro; на Vending Bench 2 он занимает первое место среди открытых моделей с симулированным балансом счёта на конец года в $4 432.
Можно ли дообучить GLM-5 самому?Да — README отсылает к Slime (собственный RL-фреймворк команды GLM, v0.3.0+) и ms-swift (v4.4.0+) для дообучения методами SFT, PPO и GRPO.
Почему GLM-5.3-Flash использует другую архитектуру, чем остальные модели GLM-5?GLM-5.3-Flash обучается из новой базовой модели с гибридной архитектурой внимания sparse+linear и Manifold-Constrained Hyper-Connections (mHC), снижая стоимость обслуживания длинного контекста при сохранении точной работы с длинным контекстом; обучение шло на мультимодальном корпусе в 30 трлн токенов.
What are similar projects?MiniCPM, Hy4-preview, DeepSeek-R1, Kimi-K2, DeepSeek-V3, Qwen3.8, jeeves, Qwen3-Coder
Which topics is it in?long-context, open-weights, speculative-decoding, mixture-of-experts, reasoning, coding, attention-mechanism, vllm, reinforcement-learning, fine-tuning, large-language-models, agentic-ai
Creator
zai-org
Size
11.2 MB
Stars
7.4K
Updated
CommitOct 15
Age
8mo
Commits · 90 Days