Paired study finds hosted decision model Jev beats open-weight Laya on most agent decision points

An arXiv preprint posted on 5 October 2026 tests two of the new single-pass decision models, hosted Jev and open-weight Laya, on 11 agent decision points built from 18 public sources and more than 7,000 base cases. Jev is significantly more accurate on 9 of the 11. Neither beats chance at zero-shot model routing, and Laya changes 30% of its answers when the option order is reversed. The authors also report errors they found in their own analysis pipeline.

What it means Decision models are being added to gateways and local runtimes fast; this is early independent evidence on where they are reliable and where they are not, especially for routing.