JSON Mode

Reviewed by Proof · Updated

Definition

JSON mode is a request option that tells a model to reply with valid JSON instead of free text.

Key takeaways

  • In the OpenAI format it's "response_format": {"type": "json_object"}.
  • It aims for output that parses as JSON. It doesn't enforce a particular set of keys.
  • You still need to say in the prompt that you want JSON, and describe the fields.
  • For a fixed schema, providers offer structured outputs. See luv13's Structured Outputs.
  • Always parse and validate the reply in code.

Why use it

When code reads the model's answer, free text is fragile. The model might add "Sure! Here's your data:" before the JSON, or wrap it in a code fence. JSON mode is meant to stop that, so json.loads() or JSON.parse() works on the reply.

JSON mode vs. structured outputs

JSON modeStructured outputs
Output parses as JSONAimed forAimed for
Matches your exact schemaNoYes, where supported
How you set it{"type": "json_object"}{"type": "json_schema", ...}

Support for each varies by provider and model. For what luv13 supports, see Structured Outputs and Request Parameters.

Tips

  • Describe the shape in the prompt: field names, types, and an example.
  • Set max_tokens high enough. A reply cut off early (finish_reason: length) is broken JSON.
  • If your code gets bad JSON, retry once, then fall back or report the error.

Example

curl https://api.luv13.ai/v1/chat/completions \
  -H "Authorization: Bearer $LUV13_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "luv13/glm-5.3-flash",
    "messages": [
      {"role": "system", "content": "Reply with JSON only, shaped like {\"city\": string, \"country\": string}."},
      {"role": "user", "content": "Where is the Eiffel Tower?"}
    ],
    "response_format": {"type": "json_object"}
  }'

If luv13 or the model doesn't support response_format, the prompt instructions alone often still produce JSON, but validate it either way.