SupraLabs has released Supra-Router-51M, a lightweight model designed to route user requests to either smaller or larger language models based on the input prompt. The model contains only 51 million parameters, allowing it to operate effectively in low-latency environments.
The release includes the model weights and the training dataset, Prompt-Routing-Dataset, available on Hugging Face.
This tool enables efficient request orchestration by dynamically selecting appropriate models for specific tasks.