Table 2:

Performance on image and no image questions

Test	Image Questions	p-value	No Image Questions	p-value
GPT-4 (4)	40.7% (37.0% – 44.4%)	p< 0.001 (4 vs. G) p= 0.885 (4 vs. 4T) p< 0.001 (G vs. 4T) p= 0.821 (4o vs. 4T) p< 0.001 (4o vs. G) p= 0.956 (4o vs. 4)	59.2% (58.2% – 60.6%)	p< 0.001 (4 vs. G) p< 0.001 (4 vs. 4T) p< 0.001 (G vs. 4T) p= 0.001 (4o vs. 4T) p< 0.001 (4o vs. G) p< 0.001 (4o vs. 4)
Gemini (G)	22.2% (20.4% – 25.9%)		44.7% (44.0% – 46.1%)
GPT-4 Turbo (4T)	44.4% (40.7% – 48.1%)		62.4% (62.1% – 63.8%)
GPT-4o (4o)	44.4% (42.2% – 46.7%)		66.7% (65.7% – 67.7%)

^*Values are presented as median percentiles and 95% confidence intervals of 30 attempts at the exam.