Amazon saves millions by replacing Sonnet 4.6 with Strands Decider

712 points · 255 comments on HN · read original →

Points and comments are a snapshot, not live.

Amazon saved millions by switching from Sonnet 4.6 to Strands Decider.

According to a blog post by Strands Agents, Amazon saved millions by replacing Anthropic's Sonnet 4.6 model with their own open-source Strands Decider for certain decision-making tasks. Strands Decider is a small, specialized decision model designed for fast experimentation and local development. The post frames this as a case study of internal adoption within Amazon, highlighting cost savings and innovation potential.

What commenters are saying

Commenters debated the economics and logic of the switch. One camp noted that if Decider costs 98% less than Sonnet and Amazon saves "millions," the vendor makes only 2% of that-but another pointed out that Strands Agents is AWS's own open-source SDK, not an external startup. Others discussed the broader pattern of moving from expensive LLMs to cheaper classifiers for simple tasks, comparing it to the shift from heavy ML pipelines to zero-shot LLM classification and back again.