GLM Chat API

Use an Official Group APIPool key. The endpoint is https://app.apipool.dev/v1/chat/completions. Confirm that your model is available in the GLM catalog before calling it.

Models and protocol

This guide covers glm-5, glm-5.1, glm-5.2, glm-5.3, and glm-5.3-flash. Use the Chat Completions messages format; do not substitute these models into the quickstart's Responses example.

Non-streaming request

Set the APIPOOL_API_KEY environment variable, then run:

curl https://app.apipool.dev/v1/chat/completions \
  -H "Authorization: Bearer $APIPOOL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"glm-5.3-flash","messages":[{"role":"user","content":"Reply only: APIPOOL-OK"}],"stream":false}'

Replace model with another available model listed above. Read the answer from choices[0].message.content and check choices[0].finish_reason; HTTP 200 alone does not establish that the answer is complete. The response's usage contains metering information. Current rates are on the selected model's detail page.

Streaming request

curl -N https://app.apipool.dev/v1/chat/completions \
  -H "Authorization: Bearer $APIPOOL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"glm-5.3-flash","messages":[{"role":"user","content":"Reply only: APIPOOL-OK"}],"stream":true}'

The response is SSE. Parse each JSON data: event, concatenate choices[].delta.content, inspect the finish reason, and keep reading until data: [DONE]. Disconnecting early may leave the reply incomplete; usage already incurred is still billed. Usage may arrive in a final chunk, so do not stop reading after the first text fragment.

Input, output, and cache rates are listed on each model's detail page. Reasoning fields are not the final answer; read message.content or streaming delta.content. When a request fails, retain the X-APIPool-Request-Id response header or error.request_id from the error body for troubleshooting.