27 billion decision-making AI scores 67.52 in public evaluation

Nace AI released a technical report for Drex Max, a 27-billion-parameter decision-making model, and disclosed a public evaluation score of 67.52 on its 0.4 Decision Index. Unlike conversational AI that generates answers token by token, Drex Max reads user-supplied documents, records, and questions and assigns probabilities to predefined answer candidates in a single inference pass, supporting multiple-choice, yes/no, and ranked-scoring formats. The company says the design cuts the step of re-analyzing generated sentences. The disclosure matters as an early benchmark bid in the emerging market for decision-oriented models rather than generative chatbots. What to watch next is whether independent evaluators reproduce the 67.52 Decision Index score and how the approach compares with larger generative models on the same tasks.