image-model-selectionlisted

# Choosing an Image Model Five models across two endpoints, priced 1-9 credits depending on model and resolution. The decision is mostly about where you are in the loop and whether the output has text or a specific product in it. ## The models **Image Studio** (`/image-studio`) — generation and editing: | Model | Use for | |---|---| | `flux-2-klein-9b` | Cheapest. Volume drafts, composition exploration | | `flux-2-pro` | **Default.** The right answer for most work | | `flux-2-max` | Highest fidelity. Hero shots, print, anything going behind ad spend | | `gpt-image-2` | Also available here | **OpenAI Images** (`/ai-image-generate`): | Model | Use for | |---|---| | `gpt-image-2` | **Default here.** Strong prompt adherence, better at text | | `dall-e-3` | Legacy. Rarely the right choice now | ## The two decisions that actually matter **1. Is there text in the image?** FLUX models render text poorly — garbled, misspelled, inconsistent. `gpt-image-2` is materially better at it. So: packaging with legible copy, signage, a label you need readable, a mockup with real words — use `gpt-image-2`. Everything else, FLUX. Even then: **if the text has to be exactly right, overlay it in an editor.** No model is reliable enough for a product label you'll ship. See `packaging-mockup-generation`. **2. Where are you in the loop?** ``` Explore composition flux-2-klein-9b cheap, many Iterate the look flux-2-pro default Final hero flux-2-max once ```