Amazon launches Strands Decider 2B open-source model

Amazon Web Services has launched Strands Decider 2B, a lightweight, open-source decision model developed by its Strands Labs division. Created by distinguished engineer Marc Brooker, the project was inspired by TypeSafe's Jev model. Brooker built the tool to address requests from AWS customers who needed a highly efficient way to handle step-by-step automation without the high latency and cost of fully featured frontier models. The model briefly claimed the top spot on the Jevbench benchmark for its size class before its official public release.

Built on the foundation of the Qwen3.5-2B large language model, Strands Decider 2B does not generate text. Instead, it evaluates pre-determined options to select the best path forward and outputs a confidence score for its choice. This architecture allows the model to run locally on modest hardware, making it highly accessible for developers building agentic workflows. According to Brooker, the primary challenge in refining these models is balancing decision-making speed and accuracy without losing the general-purpose knowledge inherent in the underlying LLM.

The release comes amid a surge of interest in decision-focused AI, arriving the same week OpenAI introduced a competing product. TypeSafe originally named its Jev model after economist William Stanley Jevons to highlight how cheaper intelligence could drive higher demand. However, TypeSafe chief executive officer Diogo Almeida remains skeptical of the sudden influx of clones. Almeida suggested that many new entrants represent "ML people wanting to implement a cool architecture" rather than teams focused on practical utility, asserting that his company does not yet face true competition.