AutoTrust releases JEV-27B, an open decision model for AI agents

AutoTrust AI has released JEV-27B, an Apache-2.0 open-weights decision model built on a frozen Qwen3.8-27B backbone. A 108.9M-parameter decision block answers yes/no, multiple-choice and 0–5 rating questions in one forward pass with a calibrated probability per option. Median latency is 137 ms, about 130 decisions per second on a single NVIDIA B200.

• Apache-2.0 license: weights, decision adapter, training and serving code, vLLM
• 108.9M-parameter block (~0.4% of the model) on frozen Qwen3.8-27B
• 137 ms median latency, ~130 decisions/sec on one B200
• Equal-weight mean 84.07% vs 83.85% for Jev 1.13 on benchmarks