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_character_consistency: bool
• photoreal_art_directed: bool
• batch_count: int (default 1)
• subjective_taste: bool
• instruction_heavy: bool

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

1. budget_per_image < 0.02 → "gpt-image-1-mini"
2. needs_text_in_image → "ideogram-3"
3. needs_character_consistency → "flux-kontext-pro"
4. photoreal_art_directed → "flux-1.1-pro"
5. batch_count >= 10 → "nano-banana"
6. subjective_taste → "midjourney-v7"
7. instruction_heavy → "gpt-image-1"
8. default → "imagen-4"

Four tasks run. Expected output:

poster:    ideogram-3
hero:      flux-1.1-pro
batch:     nano-banana
cheap:     gpt-image-1-mini