Drawing the Mona Lisa with GPT-5.6, Claude, Gemini, and Grok
Four models were evaluated on their ability to draw the Mona Lisa and Starry Night. The findings offer important insights for engineers.
Four vision models were tasked with drawing iconic images like the Mona Lisa and Starry Night. Using a set of colored-pencil tools, they produced 28 drawings, which were evaluated based on structural similarity (SSIM). The results provide valuable insights into the capabilities of these models.