Text-only LLMs can draw pixel art without any vision training

A developer shared prompts letting vision-free text-only LLMs draw 16x16 binary pixel art of shapes and animals, testing whether spatial concepts learned from language alone suffice. Qwen3-32B produced the most recognizable cats and dogs.

2026-10-26 ~ 2026-10-26 · 3 related posts

• I Asked 5 Text-Only LLMs to Draw Objects They've Never Seen — gabriberton · 2026-10-26
• The prompt behind making text-only LLMs draw 16×16 pixel images — gabriberton · 2026-10-26
• Text-only LLMs can draw pixel art: Qwen3-32B best at ASCII cats and smileys — gabriberton · 2026-10-26