### Check for existing issues

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

### What happened?

When setting `completion_model` in `general_settings` in `config.yaml`, I expect it to act as a **fallback default** — only used when the client request does **not** specify a model.

The [official docs (`config_settings.md`)](https://github.com/BerriAI/litellm/blob/9a71ba3505422eec7906df440e955c8dc9161a94/docs/my-website/docs/proxy/config_settings.md#L206) explicitly state:

> `completion_model` — *"The default model to use for completions when `model` is not specified in the request"*

**However**, when a client sends a request with an explicit `model` in the payload, the proxy **ignores the client's model entirely** and always routes the request to whatever `completion_model` is set to in `general_settings`. The `completion_model` value acts as a hard override rather than a fallback — any model specified by the client in the request body is silently discarded.

For example, with `completion_model: "gpt-5-nano"` configured, a request that explicitly sets `"model": "claude-haiku"` will still be routed to `gpt-5-nano`:

```json { "model": "cl…