Amazon releases Strands Decider 2B, its own decision model
Original: Amazon releases its own Jev clone as decision models flood the web
Why This Matters
Decision models represent a cost-efficient layer for agentic AI pipelines, and AWS entering open-source signals mainstream adoption pressure on frontier labs.
Amazon Web Services open-sourced Strands Decider 2B on October 1, 2026 — a small, fast decision model built atop Qen3.5-2B. Inspired by TypeSafe's Jev, it delivers calibrated choices between pre-set options for agentic workflows, offering lower latency and cost than full LLMs. The model runs locally and briefly topped Jevbench for its size class.
Amazon Web Services released Strands Decider 2B, a fully open-source decision model built on the Qen3.5-2B LLM backbone. Rather than generating free-form text, it picks from a closed set of pre-defined options and outputs a confidence score alongside each choice. AWS distinguished engineer Marc Brooker originated the project after experimenting with TypeSafe's Jev model independently; when his personal build briefly hit the top spot on the Jevbench leaderboard for its size class, Amazon's Strands Labs cleaned it up and shipped it publicly.
The release lands the same week OpenAI announced a comparable offering, signaling a rapid convergence around this model category. Brooker told TechCrunch the motivation came directly from AWS customers: agentic pipelines often don't need a full frontier LLM at every workflow step. A lightweight decider — faster, cheaper, and more reliable thanks to structured outputs and confidence scores — fills that gap. TypeSafe named its original model after economist William Stanley Jevons, betting that cheaper AI intelligence drives higher demand. Dozens of similar models have since appeared from researchers worldwide. Brooker cautioned that the real engineering challenge is pushing accuracy and calibration without eroding the model's general-purpose language understanding.