Amazon Bedrock Custom Models / Fine-Tuning

source

closed

training-code

closed

runs-on

AWS-Bedrock-only

license

Proprietary(managed AWS service, token+storage billed)

Proprietary managed customization service inside AWS Bedrock, no source, runs only on AWS against Bedrock-hosted foundation models.

• https://docs.aws.amazon.com/bedrock/latest/userguide/custom-models.html recorded 2026-09-23

  Lists supervised fine-tuning, reinforcement fine-tuning, and distillation as the three customization methods, delivered entirely inside Bedrock via console/API with no repository, download, or self-hostable engine referenced anywhere on the page. A billing note states training is charged by tokens processed "and model storage charged per month per model," a metered AWS commercial arrangement rather than any code license -- consistent with the recorded source:closed and license:Proprietary components.

AWS publishes no usage figure for Amazon Bedrock Custom Models: no jobs run and no count of customers customizing models. Nothing AWS publishes about Bedrock as a whole is specific to model customization, so the map leaves adoption unscored rather than borrow a figure from the wider platform.

• https://docs.aws.amazon.com/bedrock/latest/userguide/custom-models.html recorded 2026-09-28

  The page lists three customization methods and a billing note that training is charged by "the number of" tokens processed and "model storage charged per month per model", and reports no usage figure. Searched the body for customers, jobs run, models customized and usage counts: none.

Amazon Bedrock Custom Models covers supervised fine-tuning, reinforcement fine-tuning graded by Lambda reward functions, and distillation, but exposes no parallelism, precision or scale controls and publishes no training benchmark. It sits a step below OpenAI's fine-tuning API, which also offers vision fine-tuning and direct preference optimization on OpenAI's own models.

• https://docs.aws.amazon.com/bedrock/latest/userguide/custom-models.html recorded 2026-09-28

  The page lists "Supervised fine-tuning", "Reinforcement fine-tuning" with reward functions defined "using AWS Lambda" and "real-time metrics", and "Distillation" from a teacher model to a student model. Searched the body for parallelism, precision, benchmark and continued pre-training: none.