Perplexity Decider 27B: An Open-Weights LLM Decision and Routing Model
An open-weights specialized model called Perplexity Decider 27B has been released, fine-tuned from the Qwen 27B base model architecture. It is designed specifically for complex decision-making, workflow orchestration, and prompt routing tasks. Using a dedicated open-weights decision model enables developers to build self-hosted routing layers that direct prompts to the most suitable backend LLM. This approach can significantly reduce API costs, improve system efficiency, and enhance control over multi-model local AI workflows. The model leverages a 27-billion-parameter Qwen base, balancing strong reasoning capability with hardware requirements manageable for local deployment. It functions primarily as an orchestration layer to classify prompts, determine intent, and delegate execution to downstream models or tools.
## BACKGROUND
LLM routing is an architectural strategy where an entry model evaluates user requests to decide which specialized LLM or API should fulfill the task based on cost, speed, or quality. Qwen is a family of highly capable open-weights language models developed by Alibaba Cloud, frequently used as base models for open-source fine-tuning.