promptdojo_promptdojo_promptdojo_promptdojo_promptdojo_promptdojo_promptdojo_promptdojo_promptdojo_promptdojo_promptdojo_

The image model landscape — six families and what each is for — step 8 of 9

Write pick_image_model(task) that routes an image-generation task to the right 2026 model.

task is a dict with optional keys:

  • budget_per_image: float (default 1.0)
  • needs_text_in_image: bool
  • needs_vector: bool
  • photoreal_art_directed: bool
  • batch_count: int (default 1)
  • instruction_heavy: bool
  • google_premium: bool

Return one of the model name strings, in this priority order:

  1. needs_vector"recraftv4_1"
  2. needs_text_in_image"ideogram-v4"
  3. budget_per_image < 0.04"gemini-3.1-flash-lite-image"
  4. instruction_heavy"gpt-image-2"
  5. google_premium"gemini-3-pro-image"
  6. photoreal_art_directed"flux-2-pro"
  7. default (including high batch_count) → "gemini-3.1-flash-image"

Four tasks run. Expected output:

poster:    ideogram-v4
hero:      flux-2-pro
batch:     gemini-3.1-flash-image
cheap:     gemini-3.1-flash-lite-image

full-screen editor opens — close anytime to keep reading.