Grok Imagine 2.0 API: Pricing, Benchmarks, and How to Access It
Grok Imagine 2.0 API costs $0.025 per image on Unifically, ranks #2 on Arena image-edit, and is callable today. Pricing, specs, and how to access it.
Grok Imagine 2.0 is xAI's second-generation image model, released on August 7, 2026 as the Quality Mode on grok.com and in the Grok apps. Within nine days it climbed to #2 on the Arena image-edit leaderboard. xAI says its own API access is coming soon; on Unifically the model is already callable as xai/grok-imagine-2.0-image, at $0.025 per image with no subscription.
TL;DR: Grok Imagine 2.0 costs $0.025 per image for text-to-image or reference editing with up to 5 input images. On Arena it ranks #2 in image editing (Elo 1439) and #3 in text-to-image (Elo 1316) as of August 16, 2026, both behind GPT Image 2. The old
enable_protoggle is gone: quality is now a model ID. Callxai/grok-imagine-2.0-imageon Unifically or try it in the browser playground.
What is Grok Imagine 2.0?
Grok Imagine 2.0 replaces the quality half of the old Grok Imagine image setup. The first generation shipped one image model with an enable_pro switch; 2.0 splits that into two model IDs. Grok Imagine 1.0 Image stays the $0.01 speed model for drafts, and Grok Imagine 2.0 Image is the quality model for finals.
The 2.0 release is tuned for exactly the things first-generation image models fumbled: legible typography in dense layouts, region-level edits that leave the rest of the frame alone, and multi-image composition. In the Grok app it also powers a magic wand tool for point-and-edit region changes, segmentation-based selection, and background removal with transparent export. Those three are app-only tools today; the API surface is generation and reference editing.
One naming note: xAI calls it "Grok Imagine Image 2.0" and, in the app, "Quality Mode". The API model ID is xai/grok-imagine-2.0-image. Same model.
Grok Imagine 2.0 API pricing
Pricing on Unifically is flat per image and does not change when you attach references:
| Route | Model ID | Price |
|---|---|---|
| Grok Imagine 2.0 (quality) | xai/grok-imagine-2.0-image | $0.025/image |
| Grok Imagine 1.0 (speed) | xai/grok-imagine-image | $0.01/image |
Text-to-image and reference editing cost the same. A 5-reference composite edit is $0.025, the same as a bare prompt. There is no free tier: Unifically is pay-per-use, so you top up a balance and each generation deducts its exact cost, with no monthly fee and no credit expiry. A practical workflow is to draft at $0.01 on the speed model, then re-run the winning prompt on 2.0 for $0.025. Ten drafts plus one final is $0.125.
The full family's rates, including video, are in the Grok Imagine API pricing guide.
Arena benchmarks: #2 in image editing
On the Arena leaderboards as of August 16, 2026, Grok Imagine 2.0 sits at or near the top of both image boards:
| Board | Rank | Elo | Votes | Ahead of it |
|---|---|---|---|---|
| Image edit | #2 | 1439 | 5,931 | GPT Image 2 (1463) |
| Text-to-image | #3 | 1316 | 2,676 | GPT Image 2 (1381), MAI Image 2.6 preview (1336) |
Three honest notes on those numbers. GPT Image 2 leads both boards, and on image edit its lead is 24 Elo. At launch on August 7, xAI could claim #2 on both boards; text-to-image has since slipped to #3 as Microsoft's MAI Image 2.6 preview moved up, and the 2.0 ratings are still early, with vote counts in the low thousands against six figures for the incumbents. The generational jump is the clearer story: on image edit, 2.0 ranks #2 where Grok Imagine 1.0 ranks #16, and on text-to-image it ranks #3 against 1.0's #22.
Grok Imagine 2.0 vs 1.0
| Grok Imagine 2.0 | Grok Imagine 1.0 | |
|---|---|---|
| Price | $0.025/image | $0.01/image |
| Tuned for | Final quality, typography, precise edits | Speed, cheap iteration |
| Text-to-image | Yes | Yes |
| Reference editing | 1–5 images | 1–5 images |
| Aspect ratios (API) | 1:1, 2:3, 3:2, 9:16, 16:9 | Same five |
| Prompt expansion | upsample_prompt | upsample_prompt |
| Arena image-edit rank | #2 | #16 |
| Arena text-to-image rank | #3 | #22 |
The two models share an API shape, so switching is a one-line change to the model ID. If you used enable_pro: true on the first-generation model, that field has been removed; requests carrying it should move to xai/grok-imagine-2.0-image instead.
How to access the Grok Imagine 2.0 API
xAI's consumer apps expose 2.0 as Quality Mode, and xAI has said API access is "coming soon" on its own platform. On Unifically it is live now. POST to /v1/tasks:
curl -X POST https://api.unifically.com/v1/tasks \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "xai/grok-imagine-2.0-image",
"input": {
"prompt": "Product photo of a matte black watch on slate, soft window light",
"aspect_ratio": "3:2"
}
}'
Add "image_urls": ["..."] with 1 to 5 images to switch the same call into edit mode, and "upsample_prompt": true to let the API expand a short brief before generating. The response returns a task_id; poll GET /v1/tasks/<task_id> for the finished image URL. Full parameter docs live at docs.unifically.com/models/image/xai/grok-imagine-2.0-image.
No code needed to evaluate it: the Grok Imagine 2.0 Image page has a browser playground billed at the same per-image rate, and the Grok Imagine family page lists every Grok route with current prices.
Limitations
The magic wand, segmentation selection, and background removal from the launch are Grok app tools, not API parameters; through the API you steer edits with the prompt and references. The app's Smart Resize covers nine aspect ratios while the API exposes five. Moderation is stricter than on the 1.0 model in our experience, and some benign product-photo prompts can get flagged, so keep prompts specific and re-run if a request is refused. And 2.0 is the image generation only: video stays on the Grok Imagine 1.5 video line for now.
We will update this post when xAI ships more 2.0 routes (a 2.0 video model is the obvious next step), when its own API pricing lands, and as the Arena ratings settle past their early vote counts.
