xAI is retiring the grok-imagine-image-quality model slug on November 2, 2026, but developers relying on it are not facing a conventional hard shutdown. According to xAI's official release notes and its dedicated migration guide, requests sent to the retired slug after the deadline will automatically be served by grok-imagine-image-2.0 with quality set to low. The request and response shapes remain unchanged, while the replacement costs less per generated image.
The 60-day retirement notice began on September 2. The important operational detail is that the old slug will continue to resolve after November 2, meaning applications do not need an emergency code change simply to keep requests working. xAI nevertheless recommends explicitly migrating to grok-imagine-image-2.0 before the deadline because doing so gives developers direct control over the quality tier instead of silently inheriting the redirect's low setting.
The retirement is a redirect, not an API break
Beginning November 2, calls to grok-imagine-image-quality through both /v1/images/generations and /v1/images/edits will be fulfilled by grok-imagine-image-2.0. xAI says every parameter accepted by the retiring model remains accepted after the switch, and the API's request and response structures do not change.
This compatibility strategy reduces migration risk. Existing integrations can continue operating even if developers miss the retirement date, while teams that migrate intentionally can test the new model and select the quality behavior appropriate for each workload before production traffic is redirected automatically.
There is also a useful observability detail: xAI says the model field in API responses reports the model that actually served the request. Teams that temporarily retain the old slug can therefore use logs to verify when traffic begins resolving to 2.0.
The automatic replacement uses low quality
The most consequential part of the migration is not the slug change but the quality setting. Any request that continues to use grok-imagine-image-quality after November 2 is served by grok-imagine-image-2.0 with quality: "low".
Developers should not interpret “low” as a generic statement that the replacement is necessarily worse than the retiring model. It is the name of a compute and pricing tier within 2.0. xAI positions it as the closest cost match to the retiring model and reports leaderboard placements for the tier in its migration documentation. Those leaderboard claims are vendor-reported and should not substitute for testing on an application's own prompts and editing workloads.
For teams where fine visual detail is more important than minimum cost, grok-imagine-image-2.0 also supports medium. The current auto behavior uses low for generation and medium for editing, with billing based on the quality actually served. Developers can pin low or medium explicitly when predictable behavior matters.
The replacement is $0.01 cheaper per image at every listed resolution
xAI says grok-imagine-image-2.0 at low quality costs $0.01 less per generated image than grok-imagine-image-quality at each corresponding resolution. Current xAI pricing lists the retiring model at $0.05 for 1K, $0.06 for 1.5K and $0.07 for 2K output. The 2.0 low tier is $0.04, $0.05 and $0.06 respectively.
The difference can become material at API scale even though the per-image saving is small. One million images at the same resolution and comparable workload would represent $10,000 less in output charges from the one-cent reduction alone. That is a simple arithmetic illustration rather than a forecast of any customer's actual usage.
Image inputs are billed separately where applicable. xAI's current model documentation lists a $0.01 image-input charge for grok-imagine-image-2.0, while output pricing varies by resolution and quality.
Explicit migration unlocks capabilities hidden by the compatibility path
The automatic redirect is designed to preserve old integrations, but grok-imagine-image-2.0 exposes capabilities beyond the compatibility baseline. xAI says the newer model accepts the quality parameter, supports up to five source images for image editing and adds 21:9 and 5:2 aspect ratios.
The broader Imagine documentation also lists configurable generation controls around aspect ratio, resolution and output count. In other words, developers that merely allow the old slug to redirect preserve continuity, while developers that move to 2.0 explicitly gain access to the newer API surface.
One line can be enough to migrate
xAI describes the migration itself as a one-line change to the model field. Applications can replace grok-imagine-image-quality with grok-imagine-image-2.0 and optionally specify the desired quality tier. The same model change applies to image-editing requests.
The simplicity of the code change should not eliminate testing. Image-generation models can differ in composition, prompt adherence, typography, identity preservation and editing behavior even when their API contracts are compatible. Production teams should run representative prompt suites through the replacement and compare outputs before the deadline, particularly when generated images feed automated publishing or customer-facing workflows.
Auto quality deserves special attention
xAI changed the default quality behavior for grok-imagine-image-2.0 on August 28. When the quality parameter is omitted, the default is now auto. Auto currently selects low for image generation and medium for image editing.
This creates an important distinction between explicit migration and passive redirection. Requests that remain on the retired grok-imagine-image-quality slug are redirected specifically to low. An application that changes its model field to 2.0 but omits quality instead uses auto, which currently means low for generation but medium for editing. The two migration paths can therefore produce different pricing and compute behavior for edits.
Teams seeking deterministic costs should pin the tier rather than assuming the current behavior of auto will remain permanent. xAI's wording explicitly describes what auto “currently” serves, leaving room for the service to change its automatic selection later.
grok-imagine-image 1.0 is not being retired
The announcement applies specifically to grok-imagine-image-quality. xAI explicitly states that grok-imagine-image, the 1.0 model slug, is unaffected.
Current pricing lists that model separately at $0.02 per output image for both 1K and 2K, with a $0.002 image-input charge. That makes it a distinct product choice rather than another alias automatically swept into the November retirement.
The older grok-imagine-image-pro redirect becomes a two-stage migration
There is one less obvious dependency. xAI previously retired grok-imagine-image-pro and configured it to redirect to grok-imagine-image-quality. The November migration guide says those requests will follow the redirect chain to grok-imagine-image-2.0 at low quality once the quality slug is retired.
Applications still sending an even older grok-imagine-image-pro model name may therefore continue functioning through multiple compatibility layers. That does not make leaving the old slug in production ideal. Explicitly naming the current model makes configuration, billing expectations and future deprecation work easier to understand.
Model retirement is becoming an operational discipline for AI teams
The xAI change illustrates a broader issue with production generative AI. A model identifier can look like a stable dependency while the underlying platform evolves rapidly. Providers increasingly use aliases, redirects and compatibility layers to reduce breakage, but those conveniences can also hide changes in model behavior, quality settings and cost.
Engineering teams therefore need model lifecycle management alongside ordinary API monitoring. That means tracking deprecation notices, recording the model actually serving requests, pinning configurable behavior where consistency matters and maintaining evaluation suites that can be rerun before migrations.
For image systems, visual regression testing is particularly important because compatibility at the JSON level does not guarantee compatibility at the output level. The API can return exactly the same fields while the images themselves change in ways that matter to a brand or product.
The November 2 deadline is deliberately low-friction, but it should not be ignored
xAI has designed this retirement so that missing the deadline should not immediately break an application. The old slug keeps resolving, the request and response structures remain compatible and the redirected low-quality tier is cheaper. That is a relatively developer-friendly retirement path.
But automatic compatibility is not the same as an intentional migration. Teams that change to grok-imagine-image-2.0 before November 2 can benchmark the replacement, choose low or medium explicitly, understand editing costs and adopt the newer model's expanded capabilities rather than inheriting whatever the redirect selects.
The practical takeaway is straightforward: applications do not need a panic migration, but they should not treat the transparent redirect as a reason to do nothing. November 2 marks the point when grok-imagine-image-quality stops identifying the model actually doing the work. Updating the integration now keeps model choice, quality and cost under the developer's control.