vLLM Semantic Router releases d3-ultra, a 27B multimodal decision modelMachine translation
vLLM Semantic Router 发布 27B 多模态决策模型 d3-ultra
Automatically verified and published · Generated and evidence-checked automatically; not reviewed by a human.
Released under Apache-2.0, d3-ultra can process text, images or video and return probabilities for multiple choice, yes/no and localization questions in one call; it does not generate text. In the team’s private tests, it scored 87.2 on SR-Text but trailed its sibling d3-turbo on image localization.
This source permits summary display only.
Read at the original source