Researchers have published two independent artificial intelligence models capable of assisting with comprehensive patient management, from initial diagnosis through long-term treatment planning, matching or surpassing physician performance in controlled studies. The findings, published in Nature this week, introduce MIRA and Google’s AMIE, both designed to navigate the multifaceted demands of clinical decision-making beyond the narrow task specialization typical of current large language models. MIRA, developed by Jakob Kather and colleagues, operates within an isolated electronic health record environment. The system interacts with a simulated patient agent to gather clinical histories, then selects from over 85,000 potential diagnostic tests, interprets results, and formulates treatment plans, including prescriptions, procedures, and admissions. Evaluated against more than 500 real-world emergency department cases, MIRA achieved an 87.8 percent diagnostic accuracy rate, outperforming a panel of six cross-specialty physicians who scored 78.1 percent. Google’s AMIE, engineered by Mike Schaekermann and team, utilizes a Gemini-based architecture optimized for continuous, multi-visit reasoning. The system tracks disease progression and treatment responses across extended clinical timelines while cross-referencing patient data with current clinical practice guidelines and approved drug formularies. In virtual clinical examinations spanning 100 multi-visit scenarios across five medical specialties, AMIE demonstrated management reasoning on par with 21 primary care physicians. It further excelled in diagnostic precision, treatment accuracy, and strict adherence to established medical guidelines. On the newly established RxQA benchmark for complex medication reasoning, AMIE also outperformed human clinicians. Both systems address critical challenges in healthcare delivery by automating routine clinical reasoning and potentially mitigating regional physician shortages. However, researchers emphasize that neither model is yet ready for direct patient care. Further validation is required to establish generalizability, ensure safety in dynamic clinical environments, and integrate these tools into existing healthcare workflows. The studies mark a significant progression toward conversational AI agents capable of sustaining long-term disease management, laying groundwork for future clinical integration pending rigorous real-world testing and regulatory approval.