Imagen 3 is the model we reach for when the brief says "the image must match this prompt exactly." Google's prompt-following accuracy is genuinely best-in-class - it gets compositional details, counts of objects, and spatial relationships right where other models hand-wave.
The aesthetic is more photographic than artistic. If you want "a vibe," Midjourney still wins. If you want "three children, the leftmost holding a red balloon, at golden hour," Imagen 3 will deliver it.
At $0.04/image it's price-competitive with FLUX and cheaper than DALL-E. Access via Gemini, Vertex, and VideoCue.
Verdict: best prompt-following model in the category. 9.3/10.