AutoTrust AI Releases Open Decision Model for Self-Hosted AI Agents

SINGAPORE – AutoTrust AI on September 28, 2026, released JEV-27B, an open-weights model designed to handle frequent, structured decisions within AI agent workflows while preserving the underlying model's full generation and reasoning path. JEV-27B answers yes/no, multiple-choice, and 0-5 rating questions in a single forward pass, returning a calibrated probability for each option.

The model runs on a single NVIDIA B200 within customer-owned infrastructure, serving both fast System 1 decisions and deliberate System 2 generation from a single set of weights. The trained decision block, comprising approximately 0.4% of the full model, took about 9.2 hours to train on one NVIDIA B200.

"JEV-27B shows that this capability can run on one GPU inside a customer's own infrastructure, next to a reasoning model. For companies that cannot send every decision to a third-party API, that changes both the cost and the risk," said Daniel Tang, AutoTrust AI's chief executive. The model was evaluated across multiple public benchmarks, achieving strong performance metrics.

AutoTrust AI measured a median decision latency of 137 milliseconds and a sustained throughput of approximately 130 decisions per second on a single B200. This indicates a significant speed improvement compared to external API services, which reported longer latencies and lower throughput in comparative tests.