Open source project for evaluating two models on specific task

THinking about making a open source community where developers using different models could evaluate for different tasks. Imagine small model doing similarly like large model for a specific task.