GLM-5.3
Zhipu's current flagship and the top of the GLM ladder, pairing strong reasoning with permissive open weights. It targets agentic workloads and long-context coding while staying far cheaper than Western frontier models.
At a glance
Released
Sep 18, 2026
Provider
Zhipu AI
Context window
256,000 tokens
Max output
64,000 tokens
Modalities
Text + image → text
Knowledge cutoff
June 2026
API model ID
glm-5.3License
Open weights (MIT)
Pricing
Standard
Input$1.40 / 1M
Output$4.40 / 1M
A typical 1,500-token prompt with a 650-token answer costs ≈ $0.005 on the standard tier. Run your own prompt through the calculator →
What's new vs the previous model
Compared withGLM-5.2 (June 2026)
- Sharper multi-step reasoning and more reliable tool calling than GLM-5.2
- Stronger long-context retention across the full 256K window
- Higher latency on short prompts than the Flash tiers
Best use cases
Agentic coding
Drives multi-file refactors and tool-using agents with dependable function calls.
Long-document analysis
Reads and reasons over large codebases or contract bundles in a single pass.
Self-hosting
MIT weights let teams run the flagship entirely on their own hardware.
Strengths & limitations
Strengths
- Open weights (MIT)
- Strong agentic tool use
- Low cost for a flagship
Limitations
- Slower than Flash tiers
- Heavy to self-host at full precision
How to test it
- No-code: open chat.z.ai and pick GLM-5.3 from the model switcher — no API key needed.
- Playground: sign in to the Z.ai API console and run prompts in the workbench, then copy the generated code.
- API: create a key in the Z.ai API console and call the API (example below).
curl https://api.z.ai/api/paas/v4/chat/completions \
# OpenAI-compatible endpoint
-H "Authorization: Bearer $API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "glm-5.3",
"messages": [{"role": "user", "content": "Explain tokenization in one paragraph."}]
}'