Grok Imagine Image Quality Retirement: Cost Impact
xAI retires grok-imagine-image-quality on November 2. See the redirect behavior, live image prices, and a practical migration checklist.
By AI Pricing Guru Editorial Team
AI Pricing Guru articles are maintained by the editorial workflow behind the site: daily pricing snapshots, provider source checks, and review passes for model launches, subscription limits, and billing changes.
TL;DR
- xAI retires the grok-imagine-image-quality API slug on November 2, 2026, after a 60-day notice that began September 2.
- The old slug will still resolve, but xAI will serve grok-imagine-image-2.0 with quality set to low and report the serving model in the response.
- The replacement's low-quality output rate is cheaper at both published resolutions; the live table below carries the current values.
- grok-imagine-image 1.0 is not affected; teams should still switch explicitly so quality selection remains under their control.
Live Grok Imagine prices during the retirement window
| Model | 1K output | 2K output | Input image | Quality billing |
|---|---|---|---|---|
| Grok Imagine Image Quality | $0.05 | $0.07 | $0.01 | fixed by model |
| Grok Imagine Image 2.0 | $0.04 low / $0.06 medium | $0.06 low / $0.08 medium | $0.01 | auto: low generation / medium editing |
| Grok Imagine Image | $0.02 | $0.02 | $0.0020 | fixed by model |
USD per image. Rates are read from xAI's official Imagine rate card into the canonical pricing API. Input-image charges are added for editing requests.
xAI will retire grok-imagine-image-quality on November 2, 2026. The old model name will not immediately return an error. Instead, xAI says requests to that slug will be served by grok-imagine-image-2.0 with quality: "low".
That makes this a compatibility redirect with a price change, not a hard API shutdown. The 60-day notice began September 2. The request and response shapes stay the same, while the response’s model field identifies the model that actually served the call.
The live table above comes from our maintained xAI pricing dataset. It is generated from the same canonical data that powers our public pricing API, avoiding a second set of rates in article copy.
Grok Imagine retirement pricing impact
The replacement’s low-quality output rate is lower than the retiring route at both listed resolutions. The live table shows the current per-image and input-image rates, including Image 2.0’s low and medium quality tiers.
For budgeting, multiply the selected live output rate by expected generations, then add input-image charges, retries, rejected assets, storage, and any surrounding model calls. A lower generation rate only becomes a saving if approval rates and retry volume remain stable.
The old grok-imagine-image 1.0 route is unaffected and remains separately priced. Teams should not treat this notice as retirement of the whole Grok Imagine image API.
What happens to existing code?
Starting November 2, requests to /v1/images/generations and /v1/images/edits that still name grok-imagine-image-quality are redirected. xAI says every currently accepted parameter remains accepted after the switch. The replacement also supports a quality parameter, five source images for editing, and the 21:9 and 5:2 aspect ratios.
grok-imagine-image-pro already redirects to the retiring quality model. xAI says it will follow the same chain to Image 2.0 low.
This avoids an outage, but it can create a silent output change. A workflow that keeps the old model name gives xAI’s redirect control over quality. An explicit Image 2.0 call can choose low, medium, or auto. Auto currently uses low for generation and medium for editing, and xAI bills the quality actually served.
Should you migrate before November 2?
Yes. Change the model field to grok-imagine-image-2.0 and set quality explicitly where stable behavior matters. Use low for a cost-controlled path, medium when additional compute improves acceptance, or auto only when service-selected routing is acceptable.
Run a production-shaped canary before moving all traffic:
- Pin the same prompts, aspect ratios, resolutions, and source images across both slugs.
- Log the returned model, quality where available, latency, failures, and
cost_in_usd_ticks. - Blind-review outputs for text rendering, identity, product fidelity, composition, and safety.
- Compare cost per accepted asset, not only the advertised per-image rate.
- Alert on any old-slug traffic that remains after the migration date.
The redirect preserves interface compatibility, not output equivalence. A cheaper call is only a saving if approval rates and retry volume remain acceptable.
Labs status: image evaluation is blocked
This route is not added to the Cost-per-Task Labs leaderboard. That 49-task suite grades deterministic text answers and token costs; it cannot measure visual quality, edit fidelity, or cost per approved asset.
An honest retirement study needs fixed generation and editing fixtures, licensed reference images, pinned pre- and post-retirement routes, repeated outputs, blind acceptance labels, latency, retries, safety failures, exact billed cost, and a disclosed cost-per-accepted-image rule. Until that harness exists, a text leaderboard score would be fabricated evidence.
Buyer verdict
The November 2 change lowers the published output price and avoids a hard break, but teams should not rely on the redirect as their migration plan. Switch to Image 2.0 explicitly, select the intended quality, and verify cost per accepted result before the cutoff.
Use the xAI API pricing page for the maintained rate card and the AI image generation pricing guide for cross-provider context. If your image workflow also calls language models, compare OpenAI pricing and Google AI pricing, then model the surrounding token spend in the AI cost calculator.
Sources: xAI’s official retirement notice, migration guide, Imagine pricing, Image 2.0 model page, and Image Quality model page. Sources and prices checked September 3, 2026.