### Check for existing issues

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

### The Feature

## Description

GPT-5 models (e.g., `gpt-5.4`, `gpt-5.2`) use OpenAI's new `/v1/responses` endpoint, not the legacy `/v1/chat/completions` endpoint. Currently, LiteLLM routes all OpenAI models through the chat/completions endpoint, causing GPT-5 requests to fail.

## Expected Behavior

When using a GPT-5 model (any model containing "gpt-5"), LiteLLM should: - Route requests to `{api_base}/responses` - NOT route to `{api_base}/chat/completions`

## Actual Behavior

LiteLLM routes all OpenAI models, including GPT-5, to `/chat/completions`, resulting in errors:

``` litellm.BadRequestError: OpenAIException - The model 'openai.gpt-5.4' does not support the '/v1/chat/completions' API. ```

## Reproduction Steps

1. Configure a GPT-5 model in LiteLLM: ```yaml model_list: - model_name: gpt-5.4 litellm_params: model: openai/gpt-5.4 api_base: https://api.openai.com/v1 api_key: sk-... ```

2. Make a request to the model via LiteLLM proxy 3. Request fails because LiteLLM calls `/v1/chat/completions` instead of `/v1/responses`

## Env…