A model optimized for speed and efficiency, designed to provide fast inference at the cost of potentially lower performance on complex tasks.