Availability
Model Distillation is in private preview and available only to allowlisted organizations. To request access for your organization, use Forge or contact your CoreWeave account team.Billing
Model Distillation has no separate product fee. Usage is billed at the standard rates for the underlying services used by the workflow, including Serverless supervised fine-tuning (SFT), Serverless Inference, Agent Lens, Weights & Biases, and artifact storage.What Model Distillation provides
- Connect your application once: Keep the familiar Chat Completions interface and switch between models without changing how your application sends requests or reads responses.
- Learn from real usage: Automatically collect the examples your application already produces, so you can improve a model using behavior that matters to your customers.
- Build better training data: Choose the examples you want a model to learn from, review them, remove poor examples, and improve their answers when needed.
- Train models for your task: Create specialized models that can match the behavior of a much larger model while reducing cost and latency.
- Compare quality before deployment: Test models on the same examples, see where each one wins or fails, and make deployment decisions with evidence.
- Deploy safely: Test a new model with part of your traffic, increase its share gradually, and keep the previous model ready for rollback.
A typical customer journey
- Connect one AI feature to Model Distillation.
- Collect examples from normal product usage.
- Choose the examples and answers worth learning from.
- Train several candidate models.
- Compare them with the model you use today.
- Deploy the winner gradually and monitor the result.
API: Agent workflow
API: Agent workflow
Send the agent to the Quick Start for the full sequence. UI actions use the public Management API, while application traffic uses the Inference proxy API. Each actionable guide includes the equivalent request under API.