Claude Sonnet 5.5 and GPT-6 Astra tie at 41.7% on updated medical-record benchmark

Artificial Analysis says Claude Sonnet 5.5 (max with fallback) and GPT-6 Astra (max) tied at a 41.7% pass rate on its updated medical-record benchmark. Sonnet scored 64.8% on completeness and 60.9% on accuracy; Astra reached 42.2% and 95.6%, respectively. The evaluation uses 60 hard questions over synthetic records and a new three-model judge panel. Artificial Analysis says the revised scores are not directly comparable with earlier results.