Amazon AWS Releases Open-Source Strands Decider 2B as Low-Cost AI Agents Spur Demand
Updated
Updated · TechCrunch · Oct 1
Amazon AWS Releases Open-Source Strands Decider 2B as Low-Cost AI Agents Spur Demand
1 articles · Updated · TechCrunch · Oct 1
Summary
Strands Decider 2B is now available as an open-source AWS model built to choose among predefined options and return confidence scores, giving developers a faster, cheaper alternative to full-scale LLM calls.
The 2B-parameter model runs locally and is aimed at agentic workflow steps where customers need reliable next-action decisions rather than open-ended text generation.
Amazon said the project grew from distinguished engineer Marc Brooker’s experiment inspired by TypeSafe’s Jev, and the model briefly topped Jevbench among systems of its size before Strands Labs productized it.
The release lands the same week OpenAI unveiled a similar product, underscoring rising interest in decision models built on small LLM backbones such as Qen3.5-2B.
That rush also exposes a key challenge: improving speed, accuracy and calibration without weakening the broader language understanding that makes such models generally useful.