Decider-4B
Decider-4BA 4B text decision model for classification, routing, yes/no checks, and scoring, returning probabilities over defined options without generating text.
Model details
View repositoryDecider-4B is a dense, text-only decision model from Mapika, fine-tuned from Qwen3.5-4B-Base. It reads a state and typed questions with explicit answer options, then returns probability distributions over those options without decoding text. The model is designed to be called from software for structured decisions rather than conversation.
Use cases
Use it to route requests, classify business documents, check conditions against written policies, or score candidate actions. Define the answer options and criteria in each request, then use the returned probabilities to set application-specific thresholds.
Inputs and outputs
Decider-4B accepts a string, JSON object, or array as its state, with support for inputs up to 32K tokens in the documented runtime. Its SystemOne-compatible interface supports choice, yes/no, and scoring questions, returning structured answers rather than generated text.
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.
