xAI retired Grok Imagine Quality overnight — your bill fell a cent an image, and so did your quality

On November 2, 2026, xAI retired grok-imagine-image-quality — and if you were still calling it, you did not get an error. Your requests kept working, your bill got slightly smaller, and your images quietly dropped a quality tier. That is the whole story, and it is exactly the kind of change that slips past a monthly total and shows up in your output instead. [xAI migration note, "grok-imagine-image-quality Retirement on November 2, 2026," effective Nov 2, 2026]

The headline numbers

The vendor note states the $0.01 saving at every resolution; the per-resolution figures above are from AIToolsRecap's November 2 read of xAI's pricing. xAI's own recommendation is the giveaway: switch to grok-imagine-image-2.0 explicitly "before November 2," so you choose the quality you pay for rather than accepting the low default the redirect applies.

flowchart TD
    A["Before Nov 2<br/>grok-imagine-image-quality<br/>1K $0.05, 1.5K $0.06, 2K $0.07"]
    A --> B["Nov 2: slug retired<br/>No error, slug still resolves"]
    B --> C["Redirect default<br/>grok-imagine-image-2.0 at low<br/>1K $0.04, 1.5K $0.05, 2K $0.06"]
    B --> D["Explicit switch<br/>grok-imagine-image-2.0 at medium<br/>1K $0.06, 1.5K $0.07, 2K $0.08"]
    B --> E["Unaffected<br/>grok-imagine-image 1.0<br/>$0.02 per image"]

The per-unit math

A cent an image is invisible on a monthly invoice and obvious in a per-image average — which is why this teardown is priced per image, not per month. Output prices only; the $0.01 per-image input charge is identical on both sides, so it moves no comparison.

At volume, the "saving" and the "restoration" are mirror images:

So the honest framing is not "xAI cut image prices." It is: xAI repriced the default. Accept the redirect and you save 14–20% while dropping a quality tier. Refuse the downgrade and you pay 14–20% more than you did on October 31. The only free option in the whole table is the one nobody was forced into: grok-imagine-image 1.0 at $0.02 per image — half the new low price at 1K — if its output passes your acceptance bar.

flowchart LR
    L["Light user<br/>A few images a week<br/>→ Check one output, then pin quality"]
    H["Heavy generator<br/>Thousands of images<br/>→ The cent compounds; quality is the real cost"]
    T["Team / pipeline<br/>Slug still in production code<br/>→ Redirect is live; audit logs now"]
    L --> H --> T

What it means for you

If you're a light user — a few images a week through an app or a script — do one thing: generate one image and check it. If it looks fine, pin quality explicitly in your call so a future default change cannot move you again. If it looks worse, you now know why, and medium is a one-line fix.

If you're a heavy generator — thousands of images a month — the cent compounds, but quality is the real cost. Measure cost per accepted image, not per generated image. If low quality raises your reject-and-regenerate rate by even a fifth, the $10 saved per 1,000 1K images evaporates in re-runs. Run a pinned A/B: 100 images at low, 100 at medium, blind-label acceptance, then pick the cheaper accepted image — not the cheaper generated one.

If you run a team or a pipeline: the redirect is already live in your production code if the old slug is still there, and it arrived with no error to catch it. Check your logs for the model field on responses — xAI says it reports the serving model — and grep your codebase for grok-imagine-image-quality and grok-imagine-image-pro today. The fix is a one-line change to the model field plus an explicit quality parameter. Note the auto trap while you are there: auto serves low for generation and medium for editing, so the same call can bill two different rates depending on the endpoint.

The pattern to watch

This is xAI's fourth API change in five weeks, and the shape matches the rest of the market's November: prices move by redirect, not by announcement. OpenAI merged ChatGPT and Codex into one token pool from December (drafted November 2); DeepSeek started billing Search and Fetch at $0.75 per 1,000 calls on November 1 (drafted November 1). The bill no longer changes only when a vendor posts a new price list. It changes when a slug you already use starts resolving to something else.

Watch for the next one the same way: alert on cost per unit — per image, per accepted image, per search — not on the monthly total. A cent off an image never trips a budget alert. A quality tier dropping out from under a production pipeline does, but only after your users notice.

Every figure dated, every claim sourced. The AI Bill is an independent practice — no vendor sponsorship, no affiliate links, no paid placement.

Sources: xAI migration note, "grok-imagine-image-quality Retirement on November 2, 2026" (60-day notice from September 2, 2026; redirect to grok-imagine-image-2.0 at low; $0.01 per image cheaper at every resolution; grok-imagine-image 1.0 unaffected); AIToolsRecap, "AI News 2 Nov: A Grok Image Model Retires Today" (November 2, 2026) and APIWatch record of the same change (per-resolution figures $0.05/$0.06/$0.07 → $0.04/$0.05/$0.06, medium at $0.06/$0.07/$0.08, input $0.01 per image on both, grok-imagine-image 1.0 at $0.02).

Next: Book a free 45-minute teardown — screen-share of your last 3 months of AI spend, fixed-fee quote before anything starts.