Introduction to Decision Models
Amazon Web Services released an open source decision model inspired by TypeSafe’s Jev, with AI developers increasingly seeking intelligence that is more suited to computer automation than frontier LLMs.
Amazon’s Strands Decider 2B, released the same week OpenAI announced a similar offering, is a high-speed, low-cost way to sort between pre-decided options and deliver a measure of how confident it is in its choice. The model is fully open sourced, available now, and small enough to run locally.
Development and Origins
Amazon distinguished engineer Marc Brooker came up with the project after seeing Jev and trying to build his own take on such a model. The homebrew project was successful enough — it briefly reached the top spot on the Jevbench ranking for models of its size — that Amazon engineers cleaned it up and released it as an offering from their Strands Labs, an organization developing new tools and protocols for deploying AI agents.
Brooker says the need for a tool like this emerged in conversations with AWS customers, whose agentic workflows didn’t always require the capability or cost of a fully featured LLM all the time.
What originally piqued interest in this class of models was that they make a perfect decider for a workflow step, determining the next action based on the current state. It offers customers a workflow step structured for higher reliability through confidence scores and closed domain answers, alongside lower latency and reduced cost.
Architecture and Design
Like other decision models, Strands Decider is built on the torso of an LLM, specifically Qwen3.5-2B, but instead of generating text, it delivers calibrated choices. TypeSafe named their model Jev after economist William Stanley Jevons, invoking his theory that falling costs of computer intelligence can increase demand.
Industry Response
The fact that dozens of similar models have been produced by researchers since TypeSafe debuted its idea shows wide interest, but also raises questions about their long-term value. Brooker suggests the challenge lies in optimizing speedy decision-making without compromising intelligence.
There is a careful balance to find in pushing performance on accuracy and calibration without degrading general-purpose language understanding and core knowledge that makes models useful.
Still, he does not necessarily expect frontier labs to dominate the space, noting that within smaller markets, the cost to build interesting systems is relatively low.
Market Competitiveness
For their part, TypeSafe executives say they are keeping their heads down and improving future models.
While acknowledge the gold rush mentality, leadership expressed skepticism about competitors underestimating the difficulty of making models genuinely smart, stating that direct competition has not yet materialized.
Current batches of alternative models appear more like machine learning experimentation rather than dedicated efforts focused on making intelligence practical and useful.
