### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

### What happened When using the ChatGPT provider through LiteLLM, I am seeing two related failures: 1. POST /v1/chat/completions for chatgpt/* returns 403 Forbidden and asks to approve identity. 2. Health checks for the same models fail with BadRequestError: 400

This appears to be LiteLLM-specific, not an account/IP issue: - the same ChatGPT account works from the same IP through Codex CLI - the same ChatGPT account works from the same IP through VS Code Kilo / ChatGPT Pro provider

I reproduced this on both: - ghcr.io/berriai/litellm:v1.82.3 - ghcr.io/berriai/litellm:v1.83.3.rc.1

### Observed behavior:

- LiteLLM routes chatgpt/gpt-5.4 to https://chatgpt.com/backend-api/codex/chat/completions - that returns a Cloudflare challenge page and surfaces as 403 Forbidden - health checks for mode: responses send a plain string input, and the ChatGPT Responses endpoint rejects that with 400 {"detail":"Input must be a list"}

### Expected behavior: - chatgpt/gpt-5.4 should be routed through the ChatGPT Responses path and work via /v1/chat/completio…