AutoTrust AI releases JEV-27B open-weights decision model

AutoTrust AI has released JEV-27B, an Apache-2.0 open-weights model designed to facilitate structured decisions within AI agent workflows. The model is built upon a frozen Qwen3.8-27B backbone, adding a 108.9-million-parameter decision block that accounts for approximately 0.4% of the full model.

JEV-27B is capable of providing yes/no, multiple-choice, and 0–5 rating responses in a single forward pass, returning calibrated probabilities for each option. It is designed to run on a single NVIDIA B200 GPU, allowing companies to host the model on their own infrastructure to manage costs and security risks associated with third-party APIs.

According to AutoTrust AI, the model was trained in approximately 9.2 B200-hours. In evaluations across six public text-decision benchmark groups, the model achieved an equal-weight mean score of 84.07%.