### Check for existing issues

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

### What happened?

I'm using a custom provider (DeepInfra) to host `Qwen/Qwen3-Embedding-8B` for embedding. I've had a lot of difficulty setting up LiteLLM to properly run embeddings through DeepInfra.

The quick summary is that I don't think you can run embeddings when the provider is set to DeepInfra or `custom_openai`. It only works with `provider=openai` and the credentials provide an API base pointing to DeepInfra.

We fall through to litellm/main.py:6837: ``` else: raise LiteLLMUnknownProvider(model=model, custom_llm_provider=custom_llm_provider) ```

I think this issue shows off a few problems: * when an provider does not support embeddings, a very unhelpful error occurs * it's unclear if I should use the DeepInfra provider, the Custom OpenAI provider, or Openai-like, OpenAI Compatible APIs... turns out only OpenAI is supported

### Steps to Reproduce

1. Create a DeepInfra or Custom OpenAI provider model for embeddings using `Qwen/Qwen3-Embedding-8B` 2. Test the connection

### Relevant log output

```shell This is the output of a te…