AWS Releases Strands Decider 2B for Agent Workflow Decisions

AWS released Strands Decider 2B, an open-source decision model inspired by TypeSafe’s Jev, as developers increasingly seek intelligence better suited to computer automation than frontier LLMs. [1] It sorts among pre-decided options, returns a confidence measure, is available now, and is small enough to run locally. [1] The release came during the same week that OpenAI announced a similar offering, while TechCrunch described Amazon’s model as high-speed and low-cost. [1]

Amazon distinguished engineer Marc Brooker began the project after seeing Jev and building his own version, which briefly led the Jevbench ranking among models of its size. [1] Amazon engineers then cleaned up the project and released it through Strands Labs, which develops tools and protocols for deploying AI agents. [1]

Brooker said AWS customers’ agentic workflows did not always require the capability or cost of a fully featured LLM, while confidence scores and closed answer domains could support more reliable workflow steps. [1] Strands Decider uses the “torso” of Qen3.5-2B but produces calibrated choices instead of generated text. [1]

Strands Decider narrows an agent step from open-ended generation to a closed set of choices, with confidence scores available to the workflow. That makes its comparison with TypeSafe’s Jev and OpenAI’s similar offering more relevant than a broad frontier-model comparison: all three are framed around decision work rather than general text generation. Amazon’s local, open-source release also gives developers a concrete way to test whether selected workflow steps can avoid a fully featured LLM. What the sources don't address: how Strands Decider’s reliability, latency, and total operating cost compare with Jev and OpenAI’s offering under the same production workloads.