AWS open-sources a small model built to make AI agents choose

AWS has released Strands Decider 2B, an open-source model designed for a narrow but frequent job inside AI agents: choosing among options they have already been given. Rather than write prose like a chatbot, it returns a selection and a confidence signal—for example, which pre-approved next step an agent should take. Developers can use it for those routine forks in a workflow while leaving a larger language model to handle open-ended reasoning and writing. Because the smaller model can run locally and respond quickly, AWS is positioning it as a way to make agents cheaper and less sluggish when they need to make many bounded decisions.