Inference proxy API

The inference proxy exposes an OpenAI-compatible Chat Completions endpoint for project routing, provider credentials, parameter overrides, and training-grade traces.

https://proxy.training.coreweave.com/v1

Select the project with the OpenAI-Project: <entity>/<project> header and set model to "router" to route production traffic through that project’s weighted routing. Set model to a registered provider model reference, such as openai/gpt-5.5, to override the model within the project.

See Connect your application for Python, JavaScript, and cURL examples.

Last modified on October 26, 2026

Model Distillation

Serverless Training