AI tools·Chat models·China
GLM
Strong on China enterprise and coding. Qingyan app plus an API, aimed at long-horizon work.
Compiled through 2026-09
Fits China enterprise & on-prem deployments
- Vendor
- Zhipu AI
- Started
- 2019
- China availability
- Works in ChinaDirectly reachable from mainland China
- Pricing tier
- Free tier
- Current flagship
- GLM-4-Plus / GLM-Zero-Preview
- Website
- chatglm.cn
Evaluation, Pros & Cons
Editorial score based on public benchmarks and community feedback — not a third-party benchmark result
Key Advantages
- Tsinghua/Zhipu heritage; outstanding GLM-4-Plus and long-text zero-shot perf
- Enterprise-ready compliance, backed by CogVideoX and CodeGeeX stack
Limitations & Caveats
- Global community footprint still smaller than Qwen or DeepSeek
Overview
Zhipu grew from the academic GLM series into ChatGLM, Qingyan, and enterprise models. GLM-4 matched global chat quality; 2026’s GLM-5 line targets long-horizon tasks and coding. GLM-5.3 landed in August; Flash is a native multimodal MoE (~320B / 18B active) with ~1M context and GUI/browser/code tools. OpenAI- and Anthropic-style APIs ease migration.
Direction
Zhipu is betting on coding agents, enterprise compliance, and open multimodal weights. Flash is price and high-QPS; 5.3 is hard reasoning and security-style review. Qingyan and the open platform share the same family.
Version history
Aug 2026
LatestGLM-5.3 / 5.3-Flash
August release; coding + multimodal Flash.
View details
New capabilities
- GLM-5.3 for coding and security-review-style work
- 5.3-Flash native multimodal MoE, ~1M context
- OpenAI- and Anthropic-style APIs
What got stronger
- More willing, among China models, to take “edit the repo + watch the video”
- Cheaper to migrate existing GPT/Claude clients
Version notes
August 2026’s 5.3 is the current public pitch: flagship for hard reasoning and code, Flash for cheap multimodal. Dual API dialects help enterprise migrations.
May 2026
GLM-5
New flagship; long-context reasoning.
View details
New capabilities
- GLM-5 long-context reasoning flagship
What got stronger
- Steadier on long docs, audits, and code review
Version notes
GLM-5 retitled Zhipu around long-horizon work. Context and reasoning rose together and set up 5.3’s coding/multimodal drop.
Jul 2025
GLM-4.5
Stronger agents and code.
View details
New capabilities
- Stronger agents and code
What got stronger
- More usable for workflow automation, not only Q&A
Version notes
4.5 was the bridge to GLM-5: agents and code so Qingyan/API could sit inside internal systems.
Jan 2024
GLM-4
GPT-4-class chat and tools.
View details
New capabilities
- GLM-4 matched then-GPT-4-class chat and tools
- Qingyan app plus an open platform
What got stronger
- China enterprise buyers got a credible domestic closed flagship
Version notes
GLM-4 turned Zhipu from an open project into a model company you could contract. Tool use and Chinese office work were the pitch.
Mar 2023
ChatGLM
Early open Chinese chat model.
View details
New capabilities
- ChatGLM open chat model
- Chinese data plus academic lineage
What got stronger
- First cheap way for China teams to run a ChatGPT-like Chinese model
Version notes
ChatGLM was Zhipu turning the academic GLM series into a product, and it popularized open Chinese chat. Capability was limited; ecosystem and papers were not.