Skip to main content
. 2023 Oct 1;13:16492. doi: 10.1038/s41598-023-43436-9

Table 1.

Accuracy of ChatGPT and GPT4 on USMLE-type soft skills questions.

LLM USMLE sample exam (n = 21), n (%) AMBOSS (n = 59), n (%) Overall (n = 80), n (%)
ChatGPT 14 (66.6) 36 (61) 50 (62.5)
GPT-4 21 (100) 51 (86.4) 72 (90)