Amazon Web Services (AWS) has unveiled an open decision-making model inspired by TypeSafe's Jev, as AI developers increasingly look to specialized tools to automate computational processes instead of cumbersome, large language models.
Amazon Web Services (AWS) has unveiled an open decision-making model inspired by TypeSafe's Jev, as AI developers increasingly look to specialized tools to automate computational processes instead of cumbersome, large language models.
Amazon's new model — Strands Decider 2B — is fully open source, available for use now, and small enough to run locally.
The idea for the project came from Amazon lead engineer Mark Brooker, who created his own version after meeting Jev. The home-made project was so successful—it briefly ranked first in the Jevbench rankings for models of its size—that Amazon engineers refined it and released it as an official Strands Labs product.
According to Bruker, the need for such a tool emerged from conversations with AWS customers. Their autonomous workflows don't always require the high cost or full power of standard LLMs.
“What initially interested me about this class of models is that they are ideally suited to role-playing decisions at every step of the process: ‘What is the next action, given where I am?’ This gives customers the ability to structure their workflow more reliably — thanks to confidence scores and a closed response domain — and with less latency and potentially lower cost,” Mark Brooker told TechCrunch.
Like other decision-making models, Strands Decider is built on the LLM framework (in this case Qwen3.5-2B), but instead of generating text, it outputs calibrated choices.
TypeSafe named its original model Jev after economist William Stanley Jevons, alluding to his theory that a drop in the cost of a resource (such as artificial intelligence) can actually increase demand for it.
The emergence of dozens of such models since the debut of TypeSafe's solution indicates high market interest, but at the same time raises questions about their long-term value. Brooker notes that the main challenge will be optimizing the speed of decision-making without losing the general intelligence of the model. He believes that advanced AI labs will not necessarily dominate this niche, since the development of small specialized models costs from a few hundred to thousands of dollars.
“I understand that people see this as a gold rush, but they may underestimate the complexity of building truly intelligent models. The current wave is more like a desire by machine learning specialists to implement interesting architectures than the work of a team deeply committed to creating practical benefits from intelligence,” said Diogo Almeida, founder and CEO of TypeSafe.
Recall that dev.ua talked about the experience of Ukrainian IT professionals in applying Jev for real products .
