vLLM Semantic Router Releases d3-ultra: A 27B Open Multimodal Decision Model

vLLM Semantic Router released d3-ultra, a 27B parameter multimodal decision model that processes inputs once and outputs probabilities for multiple question types, supporting routing, classification, and visual tasks. It features a Qwen3.5-architecture backbone with linear and full-attention layers, handles large context windows, and achieves top scores on internal benchmarks.