The QIMMA leaderboard evaluates Arabic LLMs using a quality-first approach to combat data contamination. It shifts focus from raw benchmark scores to nuanced linguistic accuracy and cultural relevance. This framework provides Hugging Face users a reliable metric to identify models that actually understand Arabic rather than those simply memorizing test sets.