Amazon releases open-source decision model Strands Decider 2B

Amazon Web Services has released Strands Decider 2B, an open-source model designed to choose among predefined options for AI workflows. It is inspired by Jev, a decision model from TypeSafe, as developers look for models better suited to computer automation than general-purpose large language models.

Strands Decider 2B is built on Qen3.5-2B. Rather than generating text, it returns choices with confidence scores. Amazon says the model is available now and small enough to run locally. The company describes it as a fast, low-cost option for workflow steps that do not always need a full-featured large language model.

Amazon distinguished engineer Marc Brooker began the project after seeing Jev. His version briefly topped the Jevbench ranking for models of its size, after which Amazon engineers prepared it for release through Strands Labs. Brooker said confidence scores and a closed set of possible answers can make such workflow steps more reliable, with potentially lower latency and cost.

Dozens of similar models have appeared since TypeSafe introduced Jev, raising questions about their value. Brooker said developers must improve decision accuracy and calibration without weakening language understanding or general-purpose knowledge. TypeSafe CEO and founder Diogo Almeida said he did not yet see real competition for his company, and cautioned that making the models genuinely capable is difficult.