AI · 2h ago
AI Image Generators Face Off in Colored Pencil Drawing Test
A new benchmark tests GPT-5.6, Claude, Gemini, and Grok on their ability to draw the Mona Lisa using colored pencils. The results show significant variation in artistic quality and adherence to prompt constraints. No model produced a fully convincing replica, highlighting limitations in current AI image generation.
Meridian48 take
This informal test underscores how far AI art still has to go in replicating human-like creativity and precision, despite rapid advances in language models.
Read the full reporting
"Drawing" the Mona Lisa with GPT-5.6, Claude, Gemini, and Grok →
Hacker News
ai-image-generationbenchmark