Kev
An open family of small decision models, sized from under a billion parameters up to 27 billion, that can run locally.
Definition
Kev is an open family of decision models created by developer Jared Palmer, published in a range of sizes from under a billion parameters up to 27 billion so you can pick the smallest one that does your job. Like other decision models it does not generate text: it scores the available options directly and returns the winner with a confidence value. Built on top of Alibaba's open Qwen models and released with open weights, the smaller versions are light enough to run on a laptop or a phone, and the community has converted them to the common local-inference formats. It is the self-hosted counterpart to the paid hosted decision-model services.