Nace AI releases drex 1.5: a 9b decision model that evaluates options instead of generating text

Nace.AI has released Drex 1.5, a 9B parameter decision model designed to evaluate options rather than generate text. The model processes a state and typed questions, returning probabilities for each option. It achieved a score of 58.08 on the public Decision Index 0.3.1, the highest score among models with under 10B parameters. The model is available on Hugging Face and through a hosted version on OpenRouter. Drex 1.5 operates by scoring a fixed set of options in a single forward pass, supporting three question types: choice, yes/no, and ordinal score. It does not sample tokens, making temperature and top_p parameters irrelevant. The model is based on MiMo-V2.6-Distill-Qwen-9B, a distilled version of Qwen 3.5, and includes a pointer head to score options from hidden states. On the Decision Index 0.3.1, Drex 1.5 outperformed Jev on 20 out of 37 benchmarks, with strong performance in Tools (75.0) and weaker results in Knowledge and Reasoning (44.6). It also scored 86.2% on JevBench compared to Jev’s 87.0%, and showed competitive results in head-to-head games against Jev.