GLM 4.6

• 🧠 Reasoning model with tool use enabled during inference
• 🆕 Z.ai's GLM upgrade over GLM-4.5 across coding and agents
• 📏 Roughly 200K-token context for long documents and code
• ⚡ Vendor reports ~30% lower token consumption versus GLM-4.5
• 🔧 Native function calling and search-based agentic workflows
• 🔒 Open weights under the permissive MIT license
• 🏢 Built by China's Z.ai (formerly Zhipu AI)

Z.ai

Z.ai, formally Knowledge Atlas Technology Joint Stock Co., Ltd., is a Chinese technology company specializing in artificial intelligence. Previously known internationally as Zhipu AI, the company rebranded to Z.ai in 2025. Its core focus is the GLM family of…

GLM 4.6 is a large reasoning-focused language model from Z.ai, the Chinese lab formerly known as Zhipu AI, released under the MIT license with open weights. It supports a context window of roughly 200K tokens, native function calling, and search-based agentic use, with reasoning ("thinking") that can be toggled on or off at request time. Within the GLM family it succeeds GLM-4.5 and precedes GLM 4.7 and the later GLM 5 and GLM 5.1 releases.

Compared with its predecessor GLM-4.5, Z.ai reports comprehensive gains in real-world coding, long-context processing, reasoning, search, writing, and agentic applications, plus a notable efficiency improvement. The provider states GLM 4.6 consumes over 30% fewer tokens on average than GLM-4.5 while using the same inference method, and that it shows stronger tool-use and search-agent performance. Z.ai evaluated it across eight public benchmarks spanning agents, reasoning, and coding.

The model targets agentic coding workflows and integrates into frameworks and coding agents through its function-calling support. Z.ai has published its test questions and agent trajectories publicly to support reproduction of its reported results.

For context on the family's later direction, Z.ai's own GLM 4.7 model card reports further coding and reasoning gains over GLM 4.6, including 73.8% on SWE-bench (a stated +5.8%) and 42.8% on Humanity's Last Exam (a stated +12.4%), positioning GLM 4.6 as the bridge between GLM-4.5 and that newer generation.