"Drawing" the Mona Lisa with GPT-5.6, Claude, Gemini, and Grok

tryai.dev

July 21, 2026

10 min read

🔥🔥🔥🔥🔥

60/100

Summary

Four AI models, GPT-5.6, Claude, Gemini, and Grok, were tasked with drawing the Mona Lisa on a blank canvas using colored pencils. Each model executed strokes, blended colors, erased, and evaluated its work to refine the drawing.

Key Takeaways

Four AI models—GPT-5.6 Sol, Claude Fable 5, Grok 4.5, and Gemini 3.6 Flash—were tested in a drawing task to reproduce the Mona Lisa and Van Gogh's Starry Night, resulting in 28 total drawings.
GPT-5.6 Sol achieved the highest structural similarity scores when compared to the target images, outperforming the other models.
Claude Fable 5 required significantly more time and resources to produce worse outputs than GPT-5.6 in the drawing task.
The drawing arena setup is open source and allows users to run their own tests with different target images or prompts.

Read original article

View story →

Coverage & discussions

Hacker News · tryai.dev

"Drawing" the Mona Lisa with GPT-5.6, Claude, Gemini, and Grok ↗

Community discussions

Hacker News discussion ↗

Historical count: 5 · 5 saved in sample · Historical sample

Analysis ready

Community Sentiment

Mixed

Positives

Grok 4.5 shows impressive performance, outpacing composer2.5 while maintaining speed — a sign of its potential in code-focused applications.
LLMs are evolving rapidly and becoming smarter and cheaper, suggesting that today's limitations could lead to innovative uses in the future.

Concerns

Many commenters find the image outputs childish and lacking depth, with models failing to grasp basic artistic concepts like shading and reflections.
The general consensus is that these models are not image generation tools, leading to skepticism about their usefulness in creative tasks.
Critics argue that the demonstration feels more like an advertisement than a genuine showcase of AI capabilities, with overall results being disappointing.