D3
D3A 27B multimodal decision model for classification, yes/no checks, and scoring across text, images, and video, with probability-based outputs.
Model details
View repositoryD3 is the flagship multimodal decision model in vLLM Semantic Router’s Decision 3.0 family. You provide text or JSON, optional images or videos, and a set of questions. The model returns choice, yes/no, and scoring decisions with probabilities, without generating free-form text.
Use cases
Use it to route support requests using messages and attachments, classify documents or visual content, check whether a condition is met, or score evidence against criteria you define. Evaluate the model against your own data before choosing decision thresholds.
Inputs and outputs
D3 accepts multiple images and videos alongside text or structured state. Every question sees the attached media, and several questions can be answered in one request. Outputs include probabilities over the allowed choices, yes/no probabilities, and scores over defined levels.
Deploy on Baseten
Use the Deploy button to start from the Baseten deployment template. Use the deployment URL and authentication instructions from your Baseten dashboard.
