The MLCR-AA score requires both completeness and accuracy. Claude Opus 5.5 (max with fallback) reaches 63.3% completeness at 37.2% accuracy…

The MLCR-AA score requires both completeness and accuracy. Claude Opus 5.5 (max with fallback) reaches 63.3% completeness at 37.2% accuracy, while GPT-6.1 Sol (max) reaches 96.7% accuracy at 40.0% completeness. Claude Sonnet 5.5 (max with fallback) and GPT-6 Astra (max) tie at the top at 41.7%, but get there very differently: Claude Sonnet 5.5 leads on completeness at 64.8% with 60.9% accuracy, while GPT-6 Astra reaches 95.6% accuracy with 42.2% completeness.