Source: Amazon Bedrock — Model Customization

AWS documentation for supervised fine-tuning, reinforcement fine-tuning and distillation on Bedrock.

Contents
  • What: vendor documentation
  • Author: Amazon Web Services
  • Date: undated (read 2026-11-14)
  • Original: Customize your model

Summary

Covers Bedrock’s three customisation methods and the models that support fine-tuning.

Key claims

  • Supervised fine-tuning on labelled examples.
  • Reinforcement fine-tuning: reward functions (e.g. AWS Lambda) score responses; prompts or existing invocation logs serve as data.
  • Distillation: pick a teacher and a student; Bedrock generates teacher responses for your prompts and fine-tunes the student.
  • Fine-tunable models include Amazon Nova (Micro/Lite/Pro/2 Lite), Claude 3 Haiku and several Llama 3.x models, in specific regions.
  • Training is billed per token processed (corpus tokens × epochs) plus monthly storage per custom model.

Feeds

Fine-tuning · Knowledge distillation

Suggest changes →