Chuizi.AI
Chuizi.AI

Responses API

Some OpenAI models only support the Responses API and cannot be called via the Chat Completions endpoint. Chuizi.AI handles this transparently: you send a standard /v1/chat/completions request, and the gateway automatically converts it to a Responses API call when the target model requires it.

You do not need to learn the Responses API format. Your existing OpenAI SDK code works as-is.

Which Models Use the Responses API

The following models are Responses API-only. They do not support the Chat Completions endpoint on upstream providers:

ModelDescription
openai/gpt-5-codexGPT-5 Codex -- optimized for code generation
openai/gpt-5.1-codexGPT-5.1 Codex -- improved code generation
openai/gpt-5.1-codex-miniGPT-5.1 Codex Mini -- faster, lighter code model
openai/gpt-5-proGPT-5 Pro -- advanced reasoning

Bare model names (e.g., gpt-5-codex, gpt-5.1-codex-mini) also work as aliases.

All other models (GPT-4o, GPT-5, o3, o4-mini, etc.) continue to use Chat Completions normally.

How the Translation Works

When you send a request to /v1/chat/completions targeting a Responses API-only model, Chuizi.AI:

  1. Converts your request: messages array becomes Responses API input items. System messages become the instructions field.
  2. Forwards to upstream: The converted request goes to the provider's Responses API endpoint.
  3. Converts the response back: The Responses API output is translated back into a standard Chat Completions response (with choices, usage, etc.).

This all happens server-side. From your perspective, the request and response look identical to any other Chat Completions call.

Next Steps