Four leading AI vision models—GPT-5.6 Sol, Claude Fable 5, Grok 4.5, and Gemini 3.6 Flash—were tested in a drawing challenge to recreate the Mona Lisa using colored pencils on July 21, 2026. The experiment involved each model working on a blank canvas with digital colored-pencil tools, aiming to replicate the famous painting with objective scoring.
The models controlled parameters such as color, tip width, and pressure to lay down strokes, blend with smudges, and erase as needed. They could view their canvas to assess progress and decide on corrections. The test included 28 drawings across two targets—the Mona Lisa and Van Gogh's Starry Night—and five open-ended prompts. The process tracked tool use, cost, and output quality to evaluate whether the models improved over time.
This drawing arena exercise builds on previous AI creative tests, including music video generation, and provides insight into how advanced AI vision models handle complex artistic tasks. The comparison highlights differences in approach and capability among the four models, contributing to understanding AI's evolving role in visual creativity and digital art production.
The full results and detailed analysis of the AI models' performance in this drawing challenge were published on tryai.dev on July 21, 2026, offering a comprehensive look at AI-driven artistic creation.