Where do my trained weights go, and how do I deploy them?

Baseten syncs checkpoints as your job saves them. With checkpointing enabled, your training job writes checkpoints to the checkpoint directory, and Baseten uploads them in the background so they’re available to browse, reuse, or deploy — you don’t have to wait for the full job to finish to see synced checkpoints.

Synced checkpoints can be deployed to production inference from the CLI, creating a dedicated deployment on Baseten. The deployment guide covers the flow, including options for optimized LLM serving with Engine Builder.

For where checkpoints live during and after training, see storage and checkpoints.