vllm-sr/d3-turbo
vllm-sr/d3-turbo is an upgraded 27B multimodal decision model. It accepts input in text or JSON format, and optionally includes images and videos. The model processes this input to answer specific questions or requests. It can instruct the user to pick one of several options, say yes or no, rate on a scale, or point to where.
Hugging Face model
Open source ↗