Drex: Open Decision Models for Fast, Stable AI Choices
arXiv — cs.CL metadata · Nischay Dhankhar, Alisher Amantay, Zhanibek Datbayev, Darkhan Akhmetov, Yash Shah, Nurmukhamed Ubaidullayev, Raman Saparkhan, Dos Baha, Abulhair Saparov · Published on Readzer
Summary
Drex is a family of four open decision models (8B, 9B, 35B with 3B active, and 27B parameters) that output probability distributions over user-supplied options, avoiding autoregressive decoding. Trained with supervised fine-tuning and reinforcement learning from adjusted feedback (RLAF), the largest model Max achieves 67.52 on Decision Index 0.4.1, outperforming Jev 1.14.0 (60.70), and is statistically tied with d3-turbo and GPT-6 Sol on Decision Bench v4. Max answers typical requests in 183 ms, compared to 2.0 s for GPT-6 Sol and 3.5 s for Claude Opus 5.5, with stable outputs under option reordering.
This is a reviewed summary. Read the complete article at the original source.
Why it matters by industry
- Artificial intelligence
For AI practitioners, Drex offers a faster, more consistent alternative to autoregressive LLMs for decision tasks, with open-source availability and competitive accuracy.