Nace AI Open-Sources Drex 1.5: A 9B Decision Model That Scores Options, Not Text

Drex 1.5 is an open-source decision model with 8.95 billion parameters that scores a fixed set of options provided to it rather than generating text. The model reads a state and typed questions, then returns a probability for every option in a single forward pass, supporting three question types: choice, yes/no, and ordinal score. It performs comparably to closed competitor models on standard benchmarks, scoring 58.08 on the Decision Index 0.3.1, and excels with long documents, reaching 93.4% accuracy on texts between 32,000 and 128,000 tokens. The model runs on a single GPU or as a smaller compressed version on consumer hardware like Apple silicon, and developers can deploy it through multiple paths including local tools and hosted APIs.