Machine Learning
Amazon SageMaker JumpStart
Amazon SageMaker JumpStart offers pretrained models, solution templates, and example notebooks that can be deployed or customized through SageMaker AI. It accelerates experimentation while preserving customer responsibility for licensing, evaluation, and endpoint operations.
Key points
- Model cards document intended use and available artifacts
- One-click deployment still creates billable SageMaker resources
- Some models permit fine-tuning or bring-your-own-script workflows
When to use it
- Prototype an open-weight embedding model
- Benchmark a candidate foundation model on a private dataset
Exam tips
- Choose JumpStart when a catalog model fits SageMaker control requirements; Bedrock exposes managed APIs without endpoint ownership
- Review licenses and delete idle endpoints after experiments