Running GP versus the Specialist: a measured model-tier scorecard ๐ฉบ Compare small vs specialist AI model performance and cost