Optimize and migrate prompts in Amazon Bedrock

Amazon Bedrock offers prompt optimization, a model migration and optimization tool that helps you get the best performance from foundation models. Amazon Bedrock provides two prompt optimization options. Simple optimization performs a fast, heuristic rewrite of a single short prompt for one model. Advanced Prompt Optimization (AdvPO) allows you to optimize your prompts for any model on Bedrock while comparing your original prompts to optimized prompts across up to 5 models simultaneously. You can use this if you are migrating to a new model or just want to get better performance on your current model. If you are changing models, select your current model as a baseline and up to 4 other models. If you are not changing models, just select your current model to see before and after optimization. The optimizer takes your prompt templates (up to 10 per job), example user inputs for variable values (up to 100 per prompt template), ground truth answers, and an evaluation metric to guide the optimization. It's even compatible with multimodal inputs such as jpg, png, or PDF. You can provide an LLM-as-a-judge rubric, a Lambda function, or short natural language steering criteria. The evaluation steers the prompt optimization. The optimizer works in an evaluation-based feedback loop to optimize the prompt and resulting model responses, and outputs the original and final prompt templates with evaluation scores, cost estimates, and latency.

Choose an optimization method