Skip to contentToken MarketDocs

JSON Mode

This guide explains how to use Token Market JSON mode so text models return output in the specified JSON format.

What Is JSON Mode?

JSON Mode uses response_format={"type":"json_object"} to ask the model to return JSON. It is not a global switch that works for every model and channel.

Important: The current Model Market shows model types, input/output modalities, context, pricing, and channel information, but it has no dedicated JSON Mode support field. Do not treat a provider list, model tag, or text-model type as a support guarantee; validate the response with the actual model ID before using it in production.

How to Use JSON Mode

Basic Usage

Add the response_format parameter to the Chat Completions API request:

curl -X POST https://api.tokensmarket.ai/v1/chat/completions \
-H "Authorization: Bearer <API_KEY>" \
-H "Content-Type: application/json" \
-d '{
  "model": "glm-5.1",
  "messages": [
    {
      "role": "system",
      "content": "You are a professional JSON output assistant. Respond strictly in JSON format."
    },
    {
      "role": "user",
      "content": "List two colors and return them as {\"colors\":[\"color1\",\"color2\"]}"
    }
  ],
  "response_format": {"type": "json_object"}
}'

Select and Verify from Model Market

Current text models visible in Model Market include glm-5.1, deepseek-v4-flash, claude-haiku-4-5-20251001, gpt-5.4-mini, and claude-sonnet-5; models, channels, prices, and capabilities can change. For JSON Mode:

  1. In Model Market, prioritize text-generation models, then copy the real model ID shown on the current page; JSON Mode applies to text generation rather than image-only models.
  2. Send a request with response_format: {"type":"json_object"} using that model ID; do not use a provider name or a stale static list.
  3. Parse the response as JSON and validate it against your business schema; do not infer that the content matches your schema from the model name, tag, or request success alone.
  4. If the current model or channel rejects the parameter, handle the error or fall back to a normal text request with an explicit JSON instruction; otherwise switch to another text model in Model Market and validate again.

Model Market does not maintain a fixed JSON Mode support list. Do not hard-code a permanent provider or model support conclusion in the client.

Was this page helpful?