GLM parameter and protocol differences

For GLM models, build the request for your chosen protocol, then apply the adjustments on this page. These are parameter differences explicitly documented by the vendor; see Chat Completions, Responses and Messages for the full request and response structures.

Applicable models

Model	Applies to
GLM-5.3-FlashX, GLM-5.3-Flash	Chat parameter differences and multimodal additions
GLM-5.3	Chat and Responses parameter differences

GLM-5.3 text parameters

Chat Completions

Field	Difference
`thinking.type`	Thinking is always enabled; only `enabled` is supported, not `disabled`
`reasoning_effort`	Supports `low/high/max`; default: `max`
`temperature`	Range: `0～1`; default: `1`; at most two decimal places
`top_p`	Range: `0.01～1`; default: `0.95`; at most two decimal places. Adjust either this or temperature
`do_sample`	GLM extension; default: `true`. Setting `false` ignores `temperature` and `top_p`
`max_tokens`	Range: `1～131072`; the output budget includes thinking and the final answer
`tool_stream`	GLM extension; to stream function arguments, set this and `stream` to `true`
`thinking.clear_thinking`	Set `false` to retain thinking; replay the complete thinking history in its original order
`reasoning_content`	Thinking content; streamed through `delta.reasoning_content`. For interleaved-thinking tool turns, replay it unchanged along with `tool_calls` from the same message
`request_id`	GLM extension; 6–64 characters, optionally a UUID. Generated by the service if omitted
`user_id`	GLM extension; 6–128 characters. Use a stable end-user identifier without sensitive information
`tool_choice`	GLM-5.3 only supports `auto`
`response_format.type`	GLM-5.3 supports `text` or `json_object`

Responses

This section applies to GLM-5.3.

Field	Difference
`model`, `input`	Both are required
`temperature`	Range: `0～1`; default: `1`; at most two decimal places
`top_p`	Range: `0.01～1`; default: `0.95`; at most two decimal places
`max_output_tokens`	Includes thinking and the answer; default: `65536`; maximum: `131072`
`reasoning.effort`	Default: `max`; `low/medium` maps to `high`, and `xhigh` maps to `max`
`store`	Default: `false`; explicitly set `true` to save responses or maintain server-side history
`previous_response_id`	Continues a saved response; the ID is valid for seven days
`tool_choice`	Supports `none` or `auto`
`tools[].type`	Supports `function`, `namespace`, `custom` and `web_search`
`text.format.type`	Supports `text` or `json_object`

Flash / FlashX multimodal additions

Flash and FlashX share the Chat text parameters above. GLM-5.3 accepts text only and does not use these content blocks.

Input	Chat content block	Key fields
Image	`type: "image_url"`	`image_url.url` accepts an image URL or Base64 Data URL; use separate content blocks for multiple images
Video	`type: "video_url"`	`video_url.url`
File	`type: "file"`	Choose one of `file.file_id`, `file.file_url` or `file.file_data`; maximum 50 MB per file. The old `file_url` content block is retained for compatibility only

Request examples

Replace MODEL_ID with the full model ID available at your selected protocol endpoint, and use your own API Key.

This text parameter combination works with all three models.

curl 'https://api.agtcloud.ai/v1/chat/completions' \
  -H 'Authorization: Bearer your-api-key' \
  -H 'Content-Type: application/json' \
  --data '{
    "model": "MODEL_ID",
    "messages": [{"role": "user", "content": "Explain how to troubleshoot an API timeout in three points"}],
    "thinking": {"type": "enabled"},
    "reasoning_effort": "low",
    "max_tokens": 4096
  }'