Stability AI

This competition tests how well AI image models truly understand language versus how much they rely on visual habits from their training data. The prompt is deliberately simple on the surface but devilishly hard in practice. Most models default to the familiar trope of an astronaut riding a horse. By forcing the reversal, we measure three critical capabilities that separate good models from great ones:

• Strict instruction following (including negations)
• Accurate subject-object relationships and spatial hierarchy
• Resistance to strong dataset biases

While rendering simple text is no longer a real challenge for today’s SOTA models, this test shows something more interesting: how much visual taste a model has, and whether it can create a layout that feels like it came from a professional designer instead of a basic Canva template (no offense to Canva).

This challenge forces models to use one consistent handwritten style across an entire dense menu instead of defaulting to clean printed text for the smaller details, a very common failure that reveals how well they actually understand and maintain stylistic coherence.

This prompt forces models to simultaneously nail a highly specific, multi-layered commercial scene: dynamic exploded composition with multiple flying food elements, photorealistic textures, dramatic fiery lighting with embers, and precisely integrated glowing text, all while keeping strong visual impact. It is a perfect stress test that quickly separates models with true prompt mastery and creative control from those that miss details, break physics, or produce generic results.