Models
Depicta provides access to multiple AI image models. Your AI assistant (or your code) chooses which model to use for each request.
Available models
Section titled “Available models”| Model | Approx. cost | Capabilities |
|---|---|---|
| Loading models… | ||
Could not load live model data. Use depicta models or GET /v1/models to see current models and pricing.
Aliases and capabilities
Section titled “Aliases and capabilities”Use these exact aliases in the model field or with --model.
| Alias | Model | Edit | Combine | Story | Sizes | Best for |
|---|---|---|---|---|---|---|
mai-fast |
MAI-Image-2.6 Flash | Yes (1 image) | No | No | — | Default for new images; strong all-round, about 10 s |
mai |
MAI-Image-2.6 | Yes (1 image) | No | No | — | Default for graphics and diagrams; best in-image text and logos |
gemini |
Gemini 3.1 Flash Image | Yes | Yes (5) | Yes | 1K, 2K, 4K | Default for edit, combine and story; the only model with 4K |
gemini-pro |
Gemini 3 Pro Image | Yes | Yes (5) | Yes | 1K, 2K | Premium edits, most consistent relighting |
gemini-fast |
Gemini 3.1 Flash Lite Image | Yes | Yes (5) | No | 1K | Cheap drafts and simple combines; weak at small text |
seedream |
Seedream 5 Pro | Yes | Yes (5) | No | 2K | Named only: premium editorial images; slow (about a minute); processed in Malaysia by a ByteDance company; filtered content kept up to 180 days |
gpt-image |
GPT Image 2.5 Flare | Yes | Yes (5) | No | — | Named only: infographics, icon sets, exact text. OpenAI keeps inputs for up to 30 days, or longer where the law requires; avoid for photos of people. |
gpt-image-pro |
GPT Image 2.5 Sunburst | Yes | Yes (5) | No | — | Named only: photorealism; not for dense infographics. OpenAI keeps inputs for up to 30 days, or longer where the law requires; avoid for photos of people. |
“1 image” means the model accepts only the source image, with no extra reference images. Named only models run only when you or your assistant choose them by name: Depicta never uses them as a default.
Auto-selection
Section titled “Auto-selection”model is optional. Without it, Depicta uses your default model if it can handle the request, otherwise the recommended model for the operation:
| Operation | Recommended model |
|---|---|
| Generate image | mai-fast |
| Generate graphic or diagram | mai |
| Edit, combine, story | gemini |
A model that cannot take the request’s input or reference images is skipped. Depicta never picks a named-only model on its own: it runs one only when you name it, in the request or in your account preferences. The response’s model field names the model that was used. AI assistants with the Depicta skill file typically choose the model themselves based on the task and guidelines.
Explicit selection
Section titled “Explicit selection”Specify a model by alias:
depicta image "A mountain landscape" --model mai-fast{"prompt": "A mountain landscape", "model": "mai-fast"}If your account restricts which models can be used (the allowed_models preference, set in the dashboard), other models return 403 model_not_allowed.
Capabilities
Section titled “Capabilities”Not all models support all operations — see the table above. If you request an operation a model does not support, the API returns 422 model_not_capable; if a request carries more images than the model accepts, it returns 422 validation_error. Neither is charged. If the model returns no image, the API returns 451 model_content_rejected, also not charged. Check capabilities with depicta models or GET /v1/models.