Skip to main content
GLM 5.3 (Zhipu AI / Z.ai) is the text-only flagship for complex software engineering and long-horizon agents in Tess. Same 1M context as GLM 5.2 — stronger reasoning, always on, with native tools.

​
What changed vs GLM 5.2

Native reasoning and tools (function calling / MCP). No vision — for screenshots, UI, or video, use GLM 5.3 Flash (glm-5.3-flash) in the same catalog.

​
Pricing (Tess credits)

Values follow Models and Costs (credits per 100 tokens): GLM 5.3 costs about 3× GLM 5.2. Cached context reads are billed below a full input pass — keep long threads on the same model to take advantage of that.
Screenshot placeholder — model picker: Capture the chat model selector with GLM 5.3 selected.

​
Ideal use cases in Tess

  1. Long-horizon software engineering — multi-file refactors, debugging, implementation on a large codebase
  2. Persistent agents — planning, tool loops, and multi-step execution without dropping the thread
  3. Deep reasoning by default — tasks that must think before the final answer
  4. Text-only workflows when vision is not needed
  5. 1M-context sessions without switching models
Best practices
  • Reasoning cannot be turned off. Every turn spends reasoning tokens — even on low. Do not use GLM 5.3 as a cheap extractor.
  • Start on high; reserve max for planning-heavy or long tool loops.
  • Prefer GLM 5.2 (glm-5.2) when you need the same 1M context at lower cost and can live with lighter reasoning.
  • Prefer GLM 5.3 Flash when the job includes images or video, or when you want Flash-priced agents with similar agentic quality.
  • Lean on context cache in long chats to keep input cost down.
See also: GLM 5.2 · Models and Costs · GLM 5.3 (OpenRouter).