Amazon cuts costs replacing Sonnet 4.6 with Strands Decider
The gist
Amazon engineers replaced Claude Sonnet 4.6 with Strands Decider to label tickets, cutting cost per million tickets from about $22,200 to about $500.
The swap shows a cheap decision model can match a large LLM on narrow classification work.
How it unfolded
- Oct 1, 2026AWS released Strands Decider 2B as open source, the same week OpenAI announced a similar offering; it is built on a Qwen3.5-2B torso with a pointer head of about one million parameters instead of a generative head.
- Nov 9, 2026Strands Agents published a blog post detailing two Amazon engineers swapping Claude Sonnet 4.6 for Strands Decider in an internal ticketing system.
- Nov 10, 2026Coverage reported Amazon cut labeling costs by 98%, from about $22,200 to about $500 per million tickets. A Hacker News post on the story had 6 points shortly after submission.
Key terms
- LLM
- A machine learning model trained on huge amounts of text to generate, summarize, translate, and analyze language.
- open source
- Software whose source code is freely available for anyone to use, study, modify, and distribute.
Source: Wikipedia