Amazon Web Services' Strands Labs released Strands Decider 2B, an open source decision model built on a Qwen3.5-2B base, according to TechCrunch.
Decision models are built for computer automation rather than open-ended text generation, designed to make fast, reliable choices between predefined options within a closed domain, TechCrunch reported. AWS engineer Marc Brooker said the approach lets a workflow step run with lower latency and potentially lower cost than calling a full large language model, with confidence scores that make the output more reliable, according to the outlet.
Strands Decider 2B briefly reached the top of Jevbench rankings for models of its size, TechCrunch reported. It is fully open source and small enough to run locally.
The model is one of dozens that have emerged since TypeSafe introduced the first decision model, Jev, in September, TechCrunch reported, calling the pace of releases a sign of substantial developer interest in the architecture.
Every major cloud provider cloning the same narrow model category within weeks of its debut is a signal about where the margin is. Not in one big model that does everything, but in a cheap one that does a single job a thousand times a minute inside somebody else's agent loop.