# Responses (preview)

Use the Responses format for GPT models whose selected pool lists this route. Endpoint support is model-specific.

Source: https://zrelay.net/docs/api/responses/

## Request

```
curl https://api.zrelay.net/v1/responses \
  -H "Authorization: Bearer $ZRELAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-6.1-sol","input":"Hello from Zrelay","max_output_tokens":256}'
```

## Core request fields

| Field | Meaning |
| --- | --- |
| model | Exact GPT model ID with a Responses route |
| input | Text or structured Responses input |
| max_output_tokens | Output budget for this request |
| stream | Optional true for typed response events |

## SDKs and coding agents

OpenAI-compatible SDKs use `https://api.zrelay.net/v1` as their base URL. The client adds `/responses`. For Codex configuration, use the existing [setup guide](https://zrelay.net/requirements.html#codex).

## Compatibility boundary

This page does not promise every feature of another service’s Responses API. Confirm each feature on your route; do not assume server-side conversation persistence, hosted tools or background jobs are available.