Bring your own compute

Run models on infrastructure you control and expose them through the same OpenAI-compatible router - same keys, same observability, same model names. Tell us about your setup and we'll help you get connected.