Google released Gemini 3.1 Flash Lite Image, a new text-to-image model designed for speed and low cost. Simon Willison tested the tool on AI Studio by asking for a Where’s Waldo style illustration of a raccoon holding a ham radio at a woodland festival. The resulting image displayed a crowded scene with anthropomorphic animals, a Ferris wheel, and market stalls. While the visual output was dense and coherent, the model misspelled the banner text as “FOREE’S FESTIVAL” and “FOREST FIVAL”.
This update represents a practical refinement in Google’s image generation stack. The specific model name indicates a focus on efficiency for high-volume applications rather than artistic perfection. It suggests Google is prioritising deployment speed over eliminating minor spelling errors in generated text. The performance gap between this version and earlier iterations remains small enough that most users might not notice the difference in daily use.
- Model identifier is gemini-3.1-flash-lite-image
- Target use case is velocity and scale
- Spelling errors persist in generated text
