OpenAI models lead on accuracy: answering the question while citing the medical record correctly, without exaggeration or hallucination.

OpenAI models lead on accuracy: answering the question while citing the medical record correctly, without exaggeration or hallucination.
An answer is accurate when no claim in it contradicts the record, from dates, visits and measurements to the severity of a finding. OpenAI models hold the top 27 accuracy scores, led by GPT-6.1 Sol (max) at 96.7% and GPT-6 Astra (max) at 95.6%, and the best non-OpenAI model is Qwen3.8 2.4T A95B at 76.1%. Their answers stay faithful to the record but often miss key points the question asks for: GPT-6.1 Sol (max) is complete on 40.0% of questions.