xAI's image generation models — text-to-image and image editing, called with one APIMODELS key and no separate xAI account.
Authorization: Bearer YOUR_API_KEY| Model | model | Price | Notes |
|---|---|---|---|
| Grok Imagine Image | grok-imagine/grok-imagine-image | $0.03/image | Grok Imagine standard tier, text-to-image + image edit (synchronous, 19s median) |
| Grok Imagine Image Pro | grok-imagine-pro/grok-imagine-image-pro | $0.0800-$0.1200/image | Pro detail, native 1K/2K (2K default), text-to-image + multi-image edit (up to 3 refs, no extra charge) |
| Grok Imagine Image 2.0 | grok-imagine-image-2.0 | $0.0300-$0.0600/image | New generation: four tiers by resolution x quality (1K low $0.03, 2K low $0.04, 1K medium $0.045, 2K medium $0.06). Legible headline and small-print typography, up to 3 reference images fused, fourteen aspect ratios including the four ultra-wide ones. Note: no transparent background or cutout — ask for alpha and it paints the checkerboard as pixels. |
| Grok 4.2 Image | grok-4.2-image | $0.0075/image | Budget tier: same model and endpoint at a quarter of the price. Native 1K and 2K at the **same flat rate** — 2K costs no extra. No multi-image reference. |
# Step 1: Create task (text-to-image)
curl -X POST https://api.apimodels.app/v1/images/generations \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "grok-imagine/grok-imagine-image",
"prompt": "A cute cat wearing a space helmet, digital art",
"aspect_ratio": "1:1",
"n": 1
}'
# Image editing (one reference image; aspect_ratio is ignored here)
curl -X POST https://api.apimodels.app/v1/images/generations \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "grok-imagine/grok-imagine-image",
"prompt": "Replace the cat with a golden retriever",
"image_url": "https://example.com/cat.png"
}'
# Multi-image editing (up to 3 references, Pro only)
curl -X POST https://api.apimodels.app/v1/images/generations \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "grok-imagine-pro/grok-imagine-image-pro",
"prompt": "Put the dog from <IMAGE_1> into the room from <IMAGE_0>",
"image_urls": ["https://example.com/room.png", "https://example.com/dog.png"],
"resolution": "2k"
}'
# Step 2: Poll status
curl "https://api.apimodels.app/v1/images/generations?task_id=TASK_ID" \
-H "Authorization: Bearer YOUR_API_KEY"| Field | Required | Type | Description |
|---|---|---|---|
| model | Yes | string | grok-imagine/grok-imagine-image / grok-imagine-pro/grok-imagine-image-pro / grok-imagine-image-2.0（新一代，$0.03 起）/ grok-4.2-image（便宜档，$0.0075） |
| prompt | Yes | string | Text description or editing instructions |
| image_url | No | string | Reference image URL or base64, for image editing |
| image_urls | No | string[] | grok-imagine-pro only: multiple reference images (up to 3, at no extra charge — price depends on resolution only). Refer to them as <IMAGE_0>, <IMAGE_1> in the prompt. More than 3 returns an error rather than silently dropping any. |
| image_base64 | No | string | Base64-encoded reference image, for image editing |
| resolution | No | string | Either 1k or 2k; defaults to 2k. On grok-imagine-pro it is tiered ($0.08 at 1k, $0.12 at 2k); on grok-4.2-image the price is flat, so **2K costs no extra** ($0.0075 either way). Measured output: 2048x2048 for a square 2K request, 2816x1584 at 16:9. Note that passing an OpenAI-style size without an explicit resolution derives the tier from that size. |
| aspect_ratio | No | string | grok-imagine and grok-imagine-pro: 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 2:1, 1:2, 19.5:9, 9:19.5, 20:9, 9:20, auto (default 1:1; ignored when a reference image is provided). grok-4.2-image supports the same set minus the four ultra-wide values (19.5:9, 9:19.5, 20:9, 9:20) — switch to grok-imagine-image when you need those. |
| quality | No | string | grok-imagine-image-2.0 only: low or medium, defaulting to medium. **This field selects your price tier** — combined with resolution it gives the four tiers ($0.03 / $0.04 / $0.045 / $0.06). The other Grok image models do not accept it. |
| n | No | integer | Number of images: 1-10 (default 1; grok-imagine and grok-imagine-pro only support 1) |
| callback_url | No | string | Webhook URL called when task completes |
Try it in the Playground
POST https://api.apimodels.app/v1/images/generations with an Authorization: Bearer header. The model field takes grok-imagine-image-2.0 (newest generation, from $0.03), grok-imagine-image-pro (high detail, 1K $0.08 / 2K $0.12), grok-imagine-image (standard, $0.03) or grok-4.2-image (cheapest, $0.0075). One APIMODELS key covers all four — you do not need an xAI account of your own.
grok-imagine-image-2.0 is priced by resolution × quality: 1K low $0.03, 2K low $0.04, 1K medium $0.045, 2K medium $0.06 — all 25% below xAI list, and 2K low 33% below. grok-imagine-image is $0.03 per image (~4s). grok-imagine-image-pro is tiered by resolution: 1K $0.08, 2K $0.12. grok-4.2-image is $0.0075 flat. Reference images never cost extra on any of them, and failed calls are never charged.
It depends on the model. On grok-imagine-image-2.0, omitting aspect_ratio makes the output follow the first input image, but an explicit value is honoured — including the four ultra-wide ratios 19.5:9, 9:19.5, 20:9 and 9:20, which we verified one at a time against the live channel. On the older grok-imagine-image and Pro channels, a supplied reference image makes aspect_ratio ignored and the output follows the input ratio.
Neither. There is no mask parameter on any of these models — to change one region, pass the original image and describe the change in words. And none of them produce an alpha channel: ask grok-imagine-image-2.0 for a transparent background and it paints the grey-and-white checkerboard as ordinary pixels, returning an RGB PNG that only looks like a cutout in a thumbnail. Run a dedicated background-removal step instead.