Zhipu AI

What it is. Zhipu AI is the company behind the GLM model family, sold internationally as Z.ai and in China as BigModel. Why it matters. Zhipu ships genuinely open Apache-2.0 weights (GLM-5.3, GLM-5.3-Flash, GLM-5.2) while pairing them with an unusually broad agent surface (AutoGLM, GLM Coding Plan). Key characteristics. GLM-5.3 is a 744B/40B-active MoE with 1M-token context and always-on reasoning; GLM-5.3-Flash was the first open frontier model combining sparse and linear attention. What a professional should know. The Apache-2.0 labeling comes from GitHub metadata — the README has no separate weights-license section, so verify the per-model Hugging Face card before commercial reuse.

Zhipu AI (智谱) develops the GLM model family, sold internationally as Z.ai and in China as BigModel (open.bigmodel.cn). The current flagship is GLM-5.3 (released 2026-08-18, 1M context, open weights on Hugging Face), alongside the multimodal coding model GLM-5.3-Flash/FlashX (2026-08-26).

The international operator is JINGSHENG HENGXING TECHNOLOGY PTE.LTD (Singapore); the China platform is run by 北京智谱华章科技股份有限公司. The company also offers the GLM Coding Plan subscription and the Open-AutoGLM phone-agent framework. A founding date is not stated on the official pages fetched.

Why it matters

China AI Hub analysis: Zhipu AI’s structural role is the open-weight-plus-agents combination — it is the only major lab pairing Apache-2.0 open weights with a first-party phone-use agent (AutoGLM) and a coding-plan product that routes third-party tools (Claude Code, Codex, Cursor, OpenClaw) onto GLM models, which positions it as the strongest open alternative for builders who want both downloadable weights and a managed agent layer. China AI Hub analysis indicates Zhipu’s distinctive bet is breadth of surface — models, phone agents, coding plans, image/video — unified by the GLM family, with openness (Apache-2.0) as the shared thread. China AI Hub analysis indicates Zhipu’s Apache-2.0 weights plus its first-party agent surfaces make it the only lab in this database offering both downloadable models and a managed agent layer under one roof — the combination neither the closed labs nor the neutral platforms provide. GLM models are also consumed by the model-neutral platforms Dify, FastGPT and MetaGPT, which can pair Zhipu’s open weights with any workflow.

Entity hub

Models

• GLM-5.3
• GLM-5.3-Flash
• GLM-5.3-FlashX
• GLM-5.2

Products

• GLM Coding Plan
• AutoGLM
• BigModel platform

API

• Z.ai

Agents

• AutoGLM
• GLM Coding Plan

Research / Technology

• AI Agents
• Computer Use
• Mixture of Experts
• Reasoning Models
• GLM and Agent-Oriented AI

Comparisons

• DeepSeek-V4.1-Flash vs GLM-5.3-Flash
• Qwen3.8-Max vs GLM-5.3

Pricing

• Zhipu AI pricing

Benchmarks

• AutomationBench
• CyberGym
• DeepSWE
• Terminal-Bench

What is uncertain

• No official funding disclosure located; IPO/funding reports are not confirmed on official channels.
• The founding date is not stated on the official pages fetched.
• The Apache-2.0 label comes from GitHub metadata; the README has no separate weights-license section.

Sources

evidence_id	source_name	source_url	source_type	published	verified	confidence	conflict
src-companies-zhipu-ai-1	Z.ai docs — GLM-5.3 model page	https://docs.z.ai/guides/llm/glm-5.3	Official documentation	—	2026-09-20	high	—
src-companies-zhipu-ai-2	Z.ai docs — GLM-5.3-Flash model page	https://docs.z.ai/guides/vlm/glm-5.3-flash	Official documentation	—	2026-09-20	high	—
src-companies-zhipu-ai-3	Z.ai release notes	https://docs.z.ai/release-notes/new-released	Official documentation	—	2026-09-20	high	—
src-companies-zhipu-ai-4	GLM-5 GitHub repository	https://github.com/zai-org/GLM-5	Official documentation	—	2026-09-20	high	—

Labels used above: Official fact (from Z.ai docs and the GLM-5 GitHub repository), Vendor-reported claim (model capabilities and pricing published by Zhipu AI), and China AI Hub analysis (our synthesis, always introduced as such).