ChatGLM is a bilingual Chinese-English chat assistant from Zhipu AI, built on the open-weight GLM model family.
Overview #
ChatGLM #
ChatGLM is a bilingual Chinese and English chat assistant from Zhipu AI, the Tsinghua University spin-out that also publishes the GLM family of large language models. Consumers meet it as the Zhipu Qingyan app and web client at chatglm.cn, where the assistant handles conversation, document reading, writing, image understanding and code, with a free daily quota and a paid membership above it. Developers meet the same models through the BigModel open platform and the international Z.ai API, which is OpenAI-compatible. What makes ChatGLM unusual among chat assistants in 2026 is that the underlying weights are public: GLM-4.5 and the GLM-5 series ship under the MIT license, so the model you talk to in the app is also one you can self-host. Zhipu's AutoGLM extends the assistant into a phone agent that navigates apps such as WeChat and Meituan to finish multi-step tasks on your behalf.
Production credibility: Built by Zhipu AI (internationally Z.ai), founded in Beijing in 2019 as a Tsinghua University spin-out and listed on the Hong Kong Stock Exchange on 8 January 2026 as stock code 2513 (Caixin; Wikipedia); Statista records about five million monthly users of the Zhipu Qingyan app in September 2024; GLM-4.5 and GLM-5 weights are published under the MIT license by the zai-org GitHub organisation and on Hugging Face; the 2023 funding round named Alibaba, Tencent, Meituan, Ant Group, Xiaomi and HongShan as investors (Wikipedia); the US Commerce Department added Zhipu to its Entity List in January 2025 (Wikipedia).
Key Features
- Bilingual Chinese and English chat built on the GLM model family, with a hybrid thinking mode for harder reasoning
- Document, image and long-text understanding for summarising, translating and question-answering over uploads
- Writing and coding assistance, from article drafts and slide outlines to multi-file code with the ZCode environment
- AutoGLM phone agent that plans and executes multi-step tasks inside apps such as WeChat, Didi and Meituan
- Image and video generation via GLM-Image and CogVideoX models exposed in the app and the API
- Open-weight GLM-4.5 and GLM-5 series under the MIT license, downloadable from Hugging Face for self-hosting
- OpenAI-compatible REST API through the BigModel platform and Z.ai, with Flash-tier models at no charge and a free token grant
Ideal Use Case
The core user is a Chinese-speaking professional or student who wants a capable assistant that reads Chinese documents natively and runs on domestic infrastructure, or a developer who needs a strong open-weight model with a hosted API behind it. ChatGLM is the pick over ChatGPT when Chinese-language quality and mainland availability matter, and over DeepSeek when you also want the consumer app, the phone agent and the multimodal generation in one account. Teams outside China handling regulated data should weigh the Entity List status and mainland data residency before adopting it.
How ChatGLM differentiates
Against DeepSeek, the other well-known open-weight Chinese lab, ChatGLM's edge is breadth of product: a polished consumer app, AutoGLM phone agents, image and video generation and a public company behind it, whereas DeepSeek is chiefly a model and a chat box. Against ChatGPT, the differentiator is that the weights are MIT-licensed, so an enterprise can run the same GLM-5 generation on its own hardware, and the hosted pricing is aimed at Chinese and cost-sensitive markets. The GLM-5 release in February 2026 reached the top open-weight slot on Artificial Analysis and LMArena's text leaderboard according to Hugging Face's write-up, which is the strongest third-party signal for the model quality. The honest trade-offs are that the consumer app is Chinese-first with English as the second language, the company sits on the US Entity List, and data is processed in mainland China, which rules it out for many Western enterprise workloads regardless of quality.
FAQ
Q: Is ChatGLM free to use? A: Yes for everyday use. The Zhipu Qingyan app and chatglm.cn offer a free daily quota of conversations, with a paid membership for heavier use and advanced features. Developers get Flash-tier models at no charge and a free token grant on the BigModel and Z.ai platforms.
Q: How does ChatGLM compare to DeepSeek? A: Both come from Chinese labs that publish open weights. DeepSeek is primarily a model and a chat interface; ChatGLM adds a full consumer app, the AutoGLM phone agent, image and video generation and an OpenAI-compatible API from a Hong Kong-listed company. Choose DeepSeek for a minimal reasoning model, ChatGLM for the broader product.
Q: Can I run the ChatGLM models on my own servers? A: Yes. GLM-4.5, GLM-4.5-Air and the GLM-5 series are released under the MIT license by the zai-org organisation on GitHub and Hugging Face, in BF16 and FP8 variants. The flagship models are large mixture-of-experts networks, so self-hosting the full GLM-5 needs a multi-GPU node; the Air and Flash sizes fit smaller setups.
Q: Does ChatGLM work in English? A: Yes. The GLM models are trained bilingually and the international Z.ai chat and API are English-first, but the chatglm.cn consumer app and its membership flow are Chinese-first. English-speaking users generally have a smoother experience on chat.z.ai.
Q: Who owns ChatGLM and is the company stable? A: ChatGLM is built by Zhipu AI, known internationally as Z.ai, a Tsinghua University spin-out founded in 2019. It listed on the Hong Kong Stock Exchange in January 2026 as stock code 2513 and earlier raised funding from Alibaba, Tencent, Meituan and Ant Group, so it is one of the better-capitalised AI labs in China.
tl;dr
ChatGLM is Zhipu AI's bilingual assistant on the open-weight GLM models, with a free consumer app, an AutoGLM phone agent and an OpenAI-compatible API. It is the strongest choice for Chinese-language work and self-hostable frontier models; Entity List status and mainland data processing limit it for Western enterprises.
Why Use ChatGLM #
User Reviews #
Similar Tools #
AI InfrastructureThinking Machines LabFrontier AI lab founded by ex-OpenAI CTO Mira Murati. $2B seed at $12B valuation, in talks for $50-60B. Building useful and safe AI.Freemium★ 4.93♥ 462 AI AgentsMeta AIMeta AI is the Llama-powered consumer AI assistant from Meta — embedded in WhatsApp, Instagram, Messenger, plus standalone meta.ai. 1B+ users.Free★ 4.92♥ 455