Model Evaluation

Run automated quality evaluation across all models using 5 test cases scored by an AI judge.