vLLM's Semantic Router team has released d3-ultra, a 27B open multimodal decision model under Apache 2.0. It writes no text: it reads text, images or video once and returns a probability for every answer, so one call covers routing, classification and visual grounding. It scored 87.2 on the team's SR-Text test, ahead of GLiDE V1.2 on 85.1. https://www.marktechpost.com/2026/10/26/vllm-semantic-router-d3-ultra-27b-multimodal-decision-model/ #AIagent #AI #GenAI #AIresearch