Image Model Support Matrix
The current externally exposed image capabilities are not limited to a single gpt-image-* model. Based on the existing scripts and gateway adaptation logic, the image stack actually covers multiple model lines, including Gemini, Jimeng Image, Doubao Seedream, Grok Image, GPT Image, and OpenAI Images-compatible models. However, their requirements for entry points, reference images, sizes, and output formats are not exactly the same.
If you only want to know “which interface should I use,” prioritize the “Recommended entry” column; if you only want to know “whether this model supports image-to-image,” prioritize the “Reference image input” column.
Model Summary Table
| Model family | Representative models | Recommended entry | Reference image input | Typical output |
|---|
| Gemini image models | gemini-3-pro-image-preview, gemini-2.5-flash-image-preview, gemini-3.1-flash-image-preview | /v1beta/models/{model}:generateContent | inlineData with multiple parts | inlineData or URL |
| gpt-image-2 | gpt-image-2 | /v1/images/generations | JSON image, supports multiple Base64 images | url or b64_json |
| gpt-image-2-pro | gpt-image-2-pro | /v1/images/generations | JSON image, supports multiple Base64 images | url or b64_json |
| Jimeng Image | jimeng-4.0, jimeng-4.5 | /v1/images/generations, /v1/images/edits, /jm | JSON image array or file upload | url or b64_json |
| Doubao Seedream | doubao-seedream-4-0-250828, doubao-seedream-4-5-251128, doubao-seedream-5-0-260128 | /v1/images/generations | JSON image, usually a Base64 array | url or b64_json |
| Grok Image | grok-4-2-image | /v1/images/generations | JSON image, usually a Base64 array | Mainly url |
| OpenAI Images-compatible | gpt-image-1, dall-e-3, dall-e-2 | /v1/images/generations, /v1/images/edits | image / mask | url or b64_json |
By Family
Gemini image models
| Item | Description |
|---|
| Recommended API | POST /v1beta/models/{model}:generateContent |
| Example models | gemini-3-pro-image-preview, gemini-2.5-flash-image-preview, gemini-3.1-flash-image-preview |
| Request style | Native Gemini contents[].parts[] |
| Reference image | Pass image binary data through inlineData |
| Aspect ratio parameter | generationConfig.imageConfig.aspectRatio |
| Size parameter | generationConfig.imageConfig.imageSize |
| Size characteristics | gemini-3-pro-image-preview supports 1K / 2K; gemini-2.5-flash-image-preview and gemini-3.1-flash-image-preview fall back to 1K |
| Details page | Gemini Image Generation API |
gpt-image-2
| Item | Description |
|---|
| Recommended API | POST /v1/images/generations |
| Example models | gpt-image-2 |
| Request style | OpenAI Images-compatible JSON |
| Reference image | image: ["<base64>", ...] |
| Output format | response_format = url or b64_json |
| Size characteristics | Primarily built-in base aspect ratios and size tiers |
| Details page | gpt-image-2 Image Generation API |
gpt-image-2-pro
| Item | Description |
|---|
| Recommended API | POST /v1/images/generations |
| Example models | gpt-image-2-pro |
| Request style | OpenAI Images-compatible JSON |
| Reference image | image: ["<base64>", ...] |
| Output format | response_format = url or b64_json |
| Size characteristics | Additional 2K and 4K high-resolution tiers on top of the base tiers |
| Details page | gpt-image-2-pro Image Generation API |
Jimeng Image
| Item | Description |
|---|
| Recommended API | POST /v1/images/generations (OpenAI-compatible), POST /v1/images/edits (edit), POST /jm (official format) |
| Example models | jimeng-4.0, jimeng-4.5 |
| Request style | OpenAI Images-compatible JSON / multipart/form-data / official-format JSON |
| Reference image | Generation: image: ["<url>", ...]; Edit: image file upload; Official format: image_urls array |
| Aspect ratio to size mapping | The size field sets the resolution, such as 1024*1024; or use ratio to set the aspect ratio |
| Resolution | resolution field: 1k, 2k, 4k |
| Output format | url or b64_json |
| Official format | Requires the query parameters Action=CVSync2AsyncSubmitTask and Version=2022-08-31 |
| Details page | Jimeng Image Overview |
Doubao Seedream
| Item | Description |
|---|
| Recommended API | POST /v1/images/generations |
| Example models | doubao-seedream-4-0-250828, doubao-seedream-4-5-251128, doubao-seedream-5-0-260128 |
| Request style | OpenAI Images-compatible JSON |
| Reference image | image: ["<base64>", ...] |
| Aspect ratio to size mapping | 16:9 -> 2560x1440, 9:16 -> 1440x2560, 1:1 -> 2048x2048, etc. |
| Output format | Prefer b64_json, but url may also be returned alone |
| Details page | Doubao Seedream Image Generation API |
Grok Image
| Item | Description |
|---|
| Recommended API | POST /v1/images/generations |
| Example models | grok-4-2-image |
| Request style | OpenAI Images-compatible JSON |
| Reference image | image: ["<base64>", ...] |
| Aspect ratio to size mapping | 16:9 -> 2560x1440, 9:16 -> 1440x2560, 1:1 -> 2048x2048, etc. |
| Output format | Mainly returns url; do not rely on b64_json |
OpenAI Images-compatible
| Item | Description |
|---|
| Recommended API | POST /v1/images/generations, POST /v1/images/edits |
| Example models | gpt-image-1, dall-e-3, dall-e-2 |
| Request style | Native OpenAI Images fields |
| Reference image | image, mask |
| Output format | url or b64_json |
| Notes | The common semantics of this model set are described in OpenAI Images Compatibility Overview, making it suitable for compatibility integration |
Aspect Ratios and Sizes
Common aspect ratios
| Aspect ratio | Common families |
|---|
1:1 | Gemini, gpt-image-2, gpt-image-2-pro, Jimeng, Seedream, Grok |
16:9 | Gemini, gpt-image-2, gpt-image-2-pro, Jimeng, Seedream, Grok |
9:16 | Gemini, gpt-image-2, gpt-image-2-pro, Jimeng, Seedream, Grok |
4:3 / 3:4 | gpt-image-2, gpt-image-2-pro, Jimeng, Seedream, Grok |
3:2 / 2:3 | gpt-image-2, gpt-image-2-pro, Jimeng, Seedream, Grok |
21:9 | Jimeng, Seedream, Grok |
gpt-image-2 base aspect ratios and size tiers
The base tiers for gpt-image-2 are as follows:
| Preset | Actual target size | Notes |
|---|
1:1 | 1024x1024 | Base tier |
4:3 | 1536x1152 | Base tier |
3:2 | 1536x1024 | Base tier |
2:3 | 1024x1536 | Base tier |
16:9 | 1920x1080 | Base tier |
9:16 | 1080x1920 | Base tier |
gpt-image-2-pro extended high-resolution tiers
On top of the base tiers, gpt-image-2-pro also exposes the following high-resolution options:
| Preset | Actual target size | Notes |
|---|
1:1(2K) | 2048x2048 | gpt-image-2-pro extended tier |
4:3(2K) | 2048x1536 | gpt-image-2-pro extended tier |
3:2(2K) | 2560x1712 | gpt-image-2-pro extended tier |
2:3(2K) | 1712x2560 | gpt-image-2-pro extended tier |
16:9(2K) | 2048x1152 | gpt-image-2-pro extended tier |
9:16(2K) | 1152x2048 | gpt-image-2-pro extended tier |
1:1(4K) | 2880x2880 | gpt-image-2-pro extended tier |
4:3(4K) | 3840x2880 | gpt-image-2-pro extended tier |
3:2(4K) | 3840x2560 | gpt-image-2-pro extended tier |
2:3(4K) | 2560x3840 | gpt-image-2-pro extended tier |
16:9(4K) | 3840x2160 | gpt-image-2-pro extended tier |
9:16(4K) | 2160x3840 | gpt-image-2-pro extended tier |
However, note that this does not mean all upstream providers natively accept these sizes. The scripts will automatically fall back to a closer official size when necessary and write the target aspect ratio into the prompt.
Recommended Reading
- Image Series Overview
- Gemini Image Generation API
- gpt-image-2 Image Generation API
- gpt-image-2-pro Image Generation API
- Jimeng Image Overview
- Doubao Seedream Image Generation API
- OpenAI Images Compatibility Overview
- OpenAI Images-Compatible Image Editing