Amazon releases its own Jev clone as decision models flood the web

Amazon Web Services has put out its own open-source decision model, Strands Decider 2B, built for a very specific job: choosing between pre-set options quickly and cheaply. The model is meant for workflows where a full frontier LLM would be overkill, and it also returns a confidence score with its answer.

The release lands the same week OpenAI announced a similar offering. That timing says a lot about where AI tooling is drifting. Not every task needs a chatty generalist. Sometimes the useful thing is a small model that can say, with some calibration, what should happen next in a workflow.

Strands Decider started as a homebrew project by Amazon distinguished engineer Marc Brooker after he saw TypeSafe’s Jev and tried building his own version. The internal project did well enough to briefly hit the top of the Jevbench ranking for models of its size, which helped turn it into an AWS release through Strands Labs, the group building tools and protocols for AI agents.

Brooker says AWS customers were already asking for this kind of thing. Their agentic workflows didn’t always need the cost or capability of a full LLM, just a reliable step to choose the next action. Strands Decider uses the “torso” of Qen3.5-2B, but instead of generating text it outputs calibrated choices. It’s designed to be low latency, lower cost, and small enough to run locally.

The larger story is that decision models are multiplying fast. Dozens have appeared since TypeSafe introduced Jev, and that raises an obvious question: how much of this is genuinely useful, and how much is the industry chasing a neat architecture? Brooker thinks the hard part is balancing speed and calibration without dulling the model’s broader usefulness. TypeSafe, meanwhile, says it is staying quiet and working on future models instead of treating this like a gold rush.