What changed
image-generate-pro is now $0.06. It covers the Venice models that fit that price: flux-2-pro, recraft-v4, seedream-v4, grok-imagine-image, qwen-image, qwen-image-2, seedream-v5-lite, and wan-2-7-text-to-image.
image-generate-ultra is now $0.22. It covers nano-banana-pro, nano-banana-2, nano-banana-2-lite, and grok-imagine-image-quality.
GPT Image 2 stays separate because its Venice cost is $0.27 per 1K image. The new gpt-image route is priced at $0.30.
New fixed-model routes
These endpoints pin the Venice model instead of asking the caller to pick a tier:
| Endpoint | Venice model | Price |
|---|---|---:|
| flux-2-pro | flux-2-pro | $0.04 |
| seedream | seedream-v4 | $0.06 |
| recraft | recraft-v4 | $0.06 |
| qwen-image | qwen-image | $0.04 |
| gpt-image | gpt-image-2 | $0.30 |
| nano-banana | nano-banana-pro | $0.22 |
| nano-banana-2-lite | nano-banana-2-lite | $0.08 |
The model endpoints return the same hosted PNG shape as the broader image generators. The difference is routing: if a buyer searches for Flux 2 Pro, Seedream, Recraft, Qwen Image, GPT Image, or Nano Banana, there is now a direct catalog entry with a price tied to the actual upstream cost.
Provider note
Generation runs through Venice. The returned file is uploaded to fal storage so callers get a stable image URL.