kapynOpen Source

Beyond the price per token: Choosing the right OpenAI model on Amazon Bedrock for your workload

An open-source harness measures cost per correct answer for OpenAI models on Amazon Bedrock. The tool evaluates cost per token, agent trajectory cost, and rubric-graded deliverable quality, offering developers a more outcome-focused metric than traditional dollars per million tokens. It helps teams choose the most efficient model for production workloads by quantifying real-world performance and cost.

AWS ML Blog·Sep 11, 2026

Opening Kapyn…