vLLM Semantic Router has released Decision 3.0, a family of five open decision models ranging from 0.8B to 27B parameters. All models are multimodal, accepting both text and image inputs, and are distributed under the permissive Apache-2.0 license.
Unlike conventional language models that produce free-form text, Decision 3.0 models answer choice, yes/no, and scoring questions in a single call. The output is a decision rather than a generated response, which makes the models suitable for semantic routing and classification tasks where speed and determinism matter.
The release is part of the vLLM Semantic Router project, and the open weights allow developers to integrate the models directly into their own pipelines. According to the announcement, the models are designed to handle both text and images without requiring a separate generation step.