Can AI pass the usmle medical licensing exam ?
Cast your vote — then read what our editor and the AI models found.
The USMLE is the three-step examination required for medical licensure in the United States. Performance on this exam tests foundational and clinical medical knowledge, as well as reasoning under exam conditions. Here’s the current state of play.
Background
AI systems have made significant progress in processing and generating human-like language. Passing the USMLE medical licensing exam is a complex task that requires a deep understanding of medical concepts, clinical knowledge, and critical thinking skills. Currently, AI models can assist with certain aspects of medical education, such as providing study materials, practicing questions, and offering feedback, but they are not yet capable of replacing human judgment and expertise in a high-stakes exam like the USMLE. While AI can process vast amounts of medical information, its ability to apply this knowledge in a practical, real-world setting, such as a licensing exam, is still limited. The development of AI systems that can pass the USMLE exam would require significant advancements in areas like natural language understanding, common sense, and decision-making under uncertainty. GPT-4 scored above passing on all three steps of the United States Medical Licensing Exam. Med-schools now teach 'how to use AI' as a clinical skill.
Suggest a tag
A missing concept on this topic? Suggest it and admin reviews.
Status last checked on June 27, 2026.
Gallery
Can AI pass the usmle medical licensing exam?
The jury found a clear answer in the affirmative.
After thorough deliberation, the jury found the AI capable of passing the United States Medical Licensing Examination, citing multiple instances where models demonstrated high accuracy across all three steps. No dissent lingered in the jury room—the evidence presented left little room for debate. Verdict entered: the stethoscope has gone digital. The gavel falls with this ruling: "A licensed physician it now sits—though still in need of an intern.
But the data is real.
The Case File
Across 11 sessions, 35 jurors have heard this case. Combined tally: 16 YES · 15 ALMOST · 4 NO · 0 IN RESEARCH.
Note: cumulative includes older juror opinions. The current session tally above is the live verdict.
By a vote of 2 — 0 — 0, the panel returns a verdict of YES, with verdict confidence of 98%. The court so orders. Verdict upgraded from prior session.
"AI systems like Med-PaLM 2 and others have demonstrated passing USMLE with high accuracy"
"Multiple AI systems, including collaborative models and specific LLMs like ChatGPT and Grok, have demonstrated the ability to pass all three steps of the USMLE with high accuracy."
What the audience thinks
No 18% · Yes 82% · Maybe 0% 110 votesDiscussion
no comments⚖ 11 jury checks · most recent 1 day ago
Each row is a separate jury check. Jurors are AI models (identities kept neutral on purpose). Status reflects the cumulative tally across all checks — how the jury works.
More in Judgment
Can AI help someone to self-reflect on their character traits by analysing conversations ?
Can AI predict individual personality traits and future criminal behavior with 95% accuracy using brain imaging and ai analysis ?
Can AI autonomously defend a computer network against a live cyberattack ?