Chat completions

Unified public entry point for Zhipu GLM chat models.

Supported models include glm-5-turbo, glm-5, glm-4.7, glm-4.7-flash, and glm-4.5-air. The same endpoint supports both text-only chat and image understanding requests that include image_url content parts.

This document exposes the following fields: model, messages, stream, thinking, temperature, top_p, max_tokens, stop, tools, and tool_choice.

POST /v1/chat/completions

cURL

text_chat

curl --request POST \ --url https://api.powertokens.ai/v1/chat/completions \ --header 'Authorization: Bearer <token>' \ --header 'Content-Type: application/json' \ --data ' { "model": "glm-5-turbo", "messages": [ { "role": "system", "content": "You are a concise and professional assistant." }, { "role": "user", "content": "Explain vector databases in three sentences." } ], "temperature": 0.7, "max_tokens": 1024 } '

200

{
  "id": "chatcmpl_zhipu_123",
  "object": "chat.completion",
  "created": 1775174400,
  "model": "glm-5",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "A vector database is a system designed to store and retrieve vector embeddings, commonly used for semantic search, recommendations, and RAG.",
        "reasoning_content": "Define the concept first, then add the main use cases."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 42,
    "completion_tokens": 31,
    "total_tokens": 73,
    "prompt_tokens_details": {
      "cached_tokens": 0
    }
  }
}

Authorizations

Authorization · string · header · required

Send Authorization: Bearer <token> in the request headers.

Body

application/json

Request body for Zhipu chat completions.

model · enum<string> · required

Model name. Supported models include glm-5-turbo, glm-5, glm-4.7, glm-4.7-flash, and glm-4.5-air.

Available options: glm-5-turbo, glm-5, glm-4.7, glm-4.7-flash, glm-4.5-air

Example:

"glm-5-turbo"

messages · object[] · required

Message list. Supports plain text messages and image input through image_url content parts. The request must include at least one message that is not system or assistant; requests made up only of system messages or only of assistant messages are rejected upstream.

Minimum array length: 1

Show child attributes

stream · boolean · default: false

Whether to enable streaming output. When true, the response content type is text/event-stream.

thinking · object

Thinking mode configuration. Applies to models that support the thinking parameter.

Show child attributes

temperature · number · default: 1

Sampling temperature, in the range [0, 1].

Required range: 0 <= x <= 1

top_p · number · default: 0.95

Nucleus sampling threshold, in the range [0.01, 1].

Required range: 0.01 <= x <= 1

max_tokens · integer

Maximum number of output tokens.

Required range: x >= 1

stop · string[]

List of stop sequences. Only a single stop sequence is currently supported.

Maximum array length: 1

tools · object[]

Tool definitions. Only the function tool shape is currently supported.

Maximum array length: 128

Show child attributes

tool_choice · enum<string> · default: auto

Tool selection strategy. This Zhipu interface currently exposes only auto.

Available options: auto

Response

200

application/json

Success. Non-streaming mode returns JSON, while streaming mode returns an SSE event stream.

Non-streaming chat completion response.

id · string

Response ID.

object · string

Object type.

Example:

"chat.completion"

created · integer<int64>

Unix timestamp in seconds.

model · string

Actual model used.

choices · object[]

Candidate outputs returned by the model.

Show child attributes

usage · object

Token usage statistics.

Show child attributes