Grok Imagine 2.0 API

AI Image

Grok Imagine 2.0 is xAI's newest image route on OmniaKey, serving the latest Grok image model through the asynchronous, per-call media API.

Read the docs Get your API key

Per successful call$0.04
OperationsText to image
View tool pricing and billing rules

Your result will appear here once you run the model.

Where it fits

Grok Imagine 2.0 is xAI's newest image route on OmniaKey, serving the latest Grok image model through the asynchronous, per-call media API.

Choose Grok Imagine 2.0 for xAI's latest image model, then compare its results with Grok Imagine and Quality at the same per-call price.

What it does

Text to image

grok-imagine-image-2.0

Turn a text prompt into a new image.

In depth

Grok Imagine 2.0 is xAI's newest image model on OmniaKey, listed as grok-imagine-image-2.0. It generates a picture from a text prompt through the asynchronous media task API; text-to-image is the only capability listed, and one request returns one image. xAI released it on 7 August 2026, and the catalog marks it available next to the older Grok Imagine.

It is the default pick in the Grok line. It won three of our four same-prompt tests, taking the poster, the product shot and the only photograph with real shallow focus. Two habits make it predictable in production: quoting every word that must appear in the picture and saying where it sits, which put all 14 of our quoted lines on the page correctly, and setting aspect_ratio instead of leaving the shape to the model.

On 26 October 2026 we ran 11 prompts with no settings at all and got four different shapes back, all near one megapixel, because the model chose the frame each time. Two runs with aspect_ratio at 16:9 both returned 1280×720 and rebuilt the layout for the wide frame rather than cropping it. One ramen prompt also added a pile of shredded mushroom that we never listed.

Posters steered by quoted text

Put each line of text in double quotes and say where it goes: title at the top, small line at the bottom. All 14 quoted lines across our prompts came out right, middle dots included. Keep one pair of quotes around each line so separators stay with the words.

Series at a fixed shape

Lock aspect_ratio for a set of thumbnails or banners so every file lands at the same proportion. 16:9 returned 1280×720 in both runs, with the layout rebuilt for the wide frame rather than cropped. Set the ratio in the first prompt and never leave it on auto mid-series.

Dishes listed ingredient by ingredient

Photograph a dish or a product by naming every item it must show, then closing with nothing else on top. Every topping we listed appeared on the ramen, and so did a mushroom we never named. Treat the list as exhaustive and read the frame back against it.

Hands-on notes

Left on the default auto ratio for 11 prompts, Grok Imagine 2.0 returned portrait 832x1248, landscape 1248x832, square 1024x1024 and 1280x720 files.
Read the test
When we put every word of text in double quotes and said where it belongs, all 14 quoted lines in our Grok Imagine 2.0 prompts came out right.
Read the test
Our Grok Imagine 2.0 runs finished between 15.6 and 41.1 seconds after the request, with most between 16 and 25 seconds.
Read the test
When we ran our Python script through OmniaKey on 18 October 2026, Grok Imagine 2.0 returned a 1248x832 JPEG after 16 seconds.
Read the test
On our overhead ramen prompt, Grok Imagine 2.0 added a pile of dark shredded mushrooms that the prompt never listed.
Read the test

Limits to plan around

Generation only: an edit request is rejected and the count must be 1, so a set of variants means several tasks rather than one call with a number.
On auto the model picks the shape. Our 11 unset runs came back portrait, landscape, square and 1280×720, which breaks a template that expects a single size.
It invents objects that were never asked for: the ramen gained a pile of shredded mushroom. Close a list with nothing else on top and check the frame against the prompt.
A prompt upsampler rewrites the text before generation and xAI returns no rewritten prompt, so you never see what the model read. Judge the picture against your own words, not against a guess.

Compare nearby models

Model	Best for	Operations	Listed price
Grok Imagine 2.0This page grok-imagine-image-2.0	New image work: the strongest layouts in the Grok line, real depth of field and a settable aspect ratio.	Text to image	$0.04 per call
Grok Imagine grok-imagine-image	Drafts and reprints that must stay in the older Grok look, at the same per-call rate.	Text to image	$0.04 per call
Nano Banana 2.1 gemini-nano-banana-2.1	Photoreal place detail and 4K frames, plus image editing that Grok does not offer here.	Text to image · Image edit	$0.019–$0.039 per call

Sources and verification

Guide updated 2026-10-27.

Call it

Media models run asynchronously. Create a task, poll it until it finishes, then download the signed asset URL.

Create, poll, download

# 1. Create the job. 202 Accepted; the body carries the task id,
#    the pre-charged price, and a poll_url. Reuse one Idempotency-Key
#    across retries so a timeout never starts a second paid job.
curl --fail-with-body https://api.omniakey.com/v1/media/tasks \
  -H "Authorization: Bearer $OMNIAKEY_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: grok-imagine-image-2.0-001" \
  -d '{
    "model": "grok-imagine-image-2.0",
    "type": "image",
    "operation": "generate",
    "input": {"prompt": "a paper boat on a quiet lake at sunset"},
    "parameters": {"size": "1024x1024", "quality": "medium"}
  }'

# 2. Poll GET /v1/media/tasks/{id} until status is
#    succeeded | failed | canceled. Progress carries percent and stage.
curl --fail-with-body https://api.omniakey.com/v1/media/tasks/TASK_ID \
  -H "Authorization: Bearer $OMNIAKEY_API_KEY"

# 3. Download assets[0].url before expires_at, or poll again for a
#    fresh signed URL. Failed and canceled jobs are refunded.
curl -sS -o out.png "$(jq -r '.assets[0].url' result.json)"

Pricing

Image jobs are billed per successful call, not per token. Failed or canceled jobs are refunded automatically.

Per successful call

$0.04per call

More xAI image models

Grok Imagine

grok-imagine-image

Grok Imagine Video 1.5

grok-imagine-video-1.5

Grok Imagine 2.0 Prompt Guide: Structure, Examples and Real Images

Eight recipes with real images: what Grok Imagine 2.0 followed, what it added and how to steer it.

Read article

Grok Imagine API: Python Guide and 3 Models Tested Side by Side

Three Grok image models, the same four prompts. Here is which to use.

Read article

Frequently asked questions

How do I call Grok Imagine 2.0 on OmniaKey?

POST to /v1/media/tasks with the model id, type "image", and operation "generate" or "edit". Reuse one Idempotency-Key across retries so a timeout never starts a second paid job.

What does Grok Imagine 2.0 cost?

Grok Imagine 2.0 bills $0.04 per call. Creation pre-charges that amount from your prepaid balance, and failed or canceled jobs are refunded automatically.

Why does the API return a task id instead of the image?

Provider latency runs from seconds to about a minute, so OmniaKey normalizes every image route into one asynchronous lifecycle. Poll GET /v1/media/tasks/{id} until the status is succeeded, then read the signed asset URL.

How do I get text right in Grok Imagine 2.0?

Put each line in double quotes and state where it belongs in the frame, such as title across the top and the date at the bottom. Every one of the 14 quoted lines in our prompts came out right, separators included. xAI's own prompt guide says nothing about text, so the quoting method is ours, not theirs.

What size are Grok Imagine 2.0 images?

Left on auto, our 11 runs came back at 832×1248, 1248×832, 1024×1024 and 1280×720, all near one megapixel. Setting aspect_ratio to 16:9 gave 1280×720 both times. The setting accepts 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, 21:9, 5:2 and auto.

Can Grok Imagine 2.0 edit an image?

Not through OmniaKey. Grok Imagine is generation only here, so an edit request is rejected and one request returns one image. xAI's own API takes edits on this model, which is a separate service. For edits, use a model with image-edit listed, such as Nano Banana 2.1 or Seedream 5.0 Pro.

How should a Grok Imagine 2.0 prompt be ordered?

Follow xAI's order: subject, details and focal point, art direction, light, colour, then composition, written as four to six sentences about one picture. Ours ran 30 to 48 words in one to three sentences and landed close to the prompt. Lead with the subject, because a prompt that opens with mood produces a scene where the subject is a stand-in.

Why did my image include something I did not ask for?

Everything you leave out, the model fills in, and it sometimes adds real-world extras such as the mushroom on our ramen. List each item you want, then finish with nothing else on top or no text anywhere. Read the result against the prompt before you blame the settings.

How is Grok Imagine 2.0 billed?

The catalog prices it per call, one charge per image, at the same rate it lists for the other Grok image models. OmniaKey bills the task at the rate on the model page, and a failed task is refunded. The current figure sits in the price table on that page.