Skip to main content
grok-imagine-image and grok-imagine-image-quality are the legacy Grok Imagine image models on LaoZhang API. They launched on August 13, 2026 under the label “Grok Imagine 2”, remain available through the Images API, and are billed per generated image.
  • Published: August 13, 2026
  • Last verified: September 24, 2026
  • Status: Active (legacy model)

​
Key facts

xAI documents additional aspect ratios and publishes its own list prices. LaoZhang API commits only to the five ratios above, and LaoZhang retail pricing is a separate billing contract from xAI list pricing. Reconcile costs against the LaoZhang console and call logs.
“Up to 10 images” is the output count set by n, not a reference-input limit. Reference limits are model-specific: Standard accepts 1–4 references and Quality accepts 1–2.

​
Pricing

Public pricing as of September 24, 2026, per successfully generated image:
  • grok-imagine-image: $0.025. It launched at $0.02 and changed on September 3, 2026.
  • grok-imagine-image-quality: $0.045, unchanged since launch.
Current prices are in the model and pricing catalog. Use console call logs as the billing record.

​
Who should care

  • Existing integrations that already use these model IDs and need predictable per-image budgets for 1K/2K batch generation;
  • Content, advertising, and commerce teams that want multiple candidates in one request;
  • Multi-reference workflows that need four references: grok-imagine-image accepts up to four, while grok-imagine-image-2.0 accepts up to three;
  • Teams migrating from GPT Image, FLUX.2, or another Images API that want OpenAI SDK compatibility.
Text, voice, and video calls are not affected. Workflows that require 4K, mask inpainting, arbitrary pixel dimensions, or reproducible seed behavior should not assume direct compatibility.

​
What to do

1

Confirm token group and price

Open console model pricing and verify that the target token can see the model ID and its current rate.
2

Choose standard or quality

Use grok-imagine-image for routine assets and three- or four-reference composition. Compare grok-imagine-image-quality for final assets or one- to two-reference precision edits.
3

Test one image first

Test the target ratio and resolution with n=1. Inspect actual dimensions, image content, response fields, and the charge before increasing batch size or concurrency.
Minimal text-to-image request. Replace the placeholder with your LaoZhang API key; the download step uses jq:
The response lists the image URL in data[0].url. The URL is temporary, so download it promptly. For Python, reference editing, multi-image composition, migration, and errors, see the Grok Imagine Image API guide.

​
Frequently asked questions

​
What are the correct model IDs for the legacy models?

Use grok-imagine-image and grok-imagine-image-quality. The launch label “Grok Imagine 2” is not part of either ID, and it does not refer to grok-imagine-image-2.0, which is a separate, newer model.

​
Is pricing per request or per image?

Per successfully generated image, so n=10 can create ten image charges. The per-image rates are in the Pricing section above; use the console call log as the billing record.

​
Did the price change after launch?

Yes. grok-imagine-image changed on September 3, 2026; the Pricing section lists the launch and current rates. grok-imagine-image-quality did not change.

​
Do 1K and 2K cost the same?

Yes, within each model under current LaoZhang pricing. Pricing may change, so recheck the console before a large batch.

​
Why does this page list only five aspect ratios?

They are the documented LaoZhang support scope for these models. xAI documents more upstream ratios, but upstream capability alone does not prove that every value works on the LaoZhang gateway.

​
Which endpoint accepts reference images?

Use /v1/images/edits with multipart/form-data:
  • Fields: image for one file, and repeated image[] for multi-image edits.
  • Limits: grok-imagine-image supports 1–4 references, and grok-imagine-image-quality supports 1–2.
Text-to-image continues to use JSON at /v1/images/generations.

​
Should I migrate to grok-imagine-image-2.0?

Not necessarily. The legacy models remain available. Compare them with grok-imagine-image-2.0 on matched prompts: 2.0 adds 15 fixed aspect ratios and Low/Medium quality but accepts at most three references, while grok-imagine-image accepts up to four.