> ## Documentation Index
> Fetch the complete documentation index at: https://docs.geekhub.mx/llms.txt
> Use this file to discover all available pages before exploring further.

# Structured Outputs

> JSON-schema guarantee. Pass response_format and receive parseable content with no validation retries.

When you include `response_format` in your body, the model guarantees the response matches the JSON schema you sent. Useful for deterministic parsing without retries. If the selected model doesn't support it, Geek Hub **rejects the request before leaving the gateway** and returns the list of compatible models.

## Example: invoice extraction

```bash theme={null}
curl -X POST https://api.geekhub.mx/v1/chat/completions \
  -H "Authorization: Bearer ghub_sk_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-4.1-mini",
    "messages": [
      {"role":"system","content":"You extract data from invoices."},
      {"role":"user","content":"Invoice #A123 from Geek Vibes for $12,345.67 dated June 12, 2026."}
    ],
    "response_format": {
      "type": "json_schema",
      "json_schema": {
        "name": "invoice",
        "strict": true,
        "schema": {
          "type": "object",
          "properties": {
            "number": { "type": "string" },
            "issuer": { "type": "string" },
            "total_usd": { "type": "number" },
            "date_iso": { "type": "string", "format": "date" }
          },
          "required": ["number", "issuer", "total_usd", "date_iso"],
          "additionalProperties": false
        }
      }
    }
  }'
```

`content` arrives as a parseable JSON string. `JSON.parse(content)` gives you the typed object directly.

## Simple mode

```json theme={null}
"response_format": { "type": "json_object" }
```

<Warning>
  OpenAI requires the prompt to contain the word "json" when using `json_object`. Not a Geek Hub requirement — provider-specific.
</Warning>

## Compatible models

12 of 17 chat models support structured outputs.

| Family    | Models                                               |
| --------- | ---------------------------------------------------- |
| Anthropic | claude-opus-4-8, claude-sonnet-4-6, claude-haiku-4-5 |
| Google    | gemini-2.5-pro, gemini-2.5-flash                     |
| OpenAI    | gpt-5, gpt-4.1, gpt-4.1-mini, o4-mini                |
| xAI       | grok-4, grok-3, grok-3-mini                          |

## Rejection

```json theme={null}
HTTP/1.1 422 Unprocessable Entity

{
  "error": {
    "type": "structured_outputs_not_supported",
    "message": "Model \"deepseek/deepseek-chat\" does not support structured outputs.",
    "model": "deepseek/deepseek-chat",
    "compatible_models": [
      "anthropic/claude-opus-4-8",
      "openai/gpt-4.1-mini",
      "..."
    ]
  }
}
```

<Tip>
  Combine with [Model Fallbacks](/en/features/fallbacks): pass a list of candidates and unsupported ones get **skipped** instead of failing.
</Tip>