AI models Claude Sonnet 5.5 and GPT-6 Astra lead medical reasoning benchmark

Anthropic’s Claude Sonnet 5.5 and OpenAI’s GPT-6 Astra have tied for the top position on the MLCR-AA medical reasoning leaderboard. According to an update from Artificial Analysis, both models achieved a 41.7% pass rate in synthesizing long medical and insurance records.

The benchmark, developed by the insurtech platform Wisedocs, utilizes synthetic case files containing between 25,000 and 64,000 tokens to test an AI's ability to perform record synthesis tasks similar to those handled by claims professionals. While the two models share the top spot, their strengths differ: Claude Sonnet 5.5 led in the completeness metric at 64.8%, whereas OpenAI’s GPT-6.1 Sol led in accuracy at 96.7%.

The median score across all models on the leaderboard remains below 15%, highlighting a significant performance gap between the leaders and the rest of the field. No model in the current update exceeded a 42% pass rate.