# grok-imagine-image-quality Retired: Now 2.0 at Low, What to Set

> xAI retired grok-imagine-image-quality on November 2, 2026. Old calls still work on grok-imagine-image-2.0 at low; set model and quality yourself.

- URL: https://blog.laozhang.ai/en/posts/grok-imagine-image-quality-retired
- Published: 2026-11-03
- Updated: 2026-11-03
- Author: LaoZhang AI Team (https://blog.laozhang.ai/en/about)
- Topic: API Guides
- Tags: Grok Imagine, xAI API, Image Generation API, Model Migration, API Pricing

---
xAI retired the `grok-imagine-image-quality` model on November 2, 2026, but code that still sends that ID does not break. On the xAI API, the old slug keeps resolving: generation and edit requests to it are now served by `grok-imagine-image-2.0` with `quality` set to `low`, the request and response shapes are unchanged, and you are billed at the 2.0 `low` rate, which at 1K is $0.04 per image instead of the old $0.05, according to [xAI's grok-imagine-image-quality migration guide](https://docs.x.ai/developers/migration/imagine-image-quality-nov-2).

What you should do now is change the `model` field to `grok-imagine-image-2.0` and pass `quality` yourself: `low` is the closest cost match to the old model, `medium` buys finer detail. Pinning it matters most for image edits, because 2.0 without a `quality` value edits at `medium` and bills at the medium rate, not the 1K `low` price. If you call the model through a gateway or a wrapper (a ComfyUI node, SillyTavern, a proxy), the gateway decides what runs and what you pay. On laozhang.ai, for example, the old ID was still callable at $0.045 per image as of November 3, 2026, while `grok-imagine-image-2.0` costs a flat $0.055, so check which model ID your provider actually bills.

*Disclosure: [laozhang.ai is our own service](https://blog.laozhang.ai/en/about#laozhang). Its prices are platform pricing, not an independent comparison.*

If you came here because of the Grok app's Quality Mode, that is a different thing from this API model ID, and nothing published says it changed on November 2; the [section on the app](#grok-app-quality-mode-is-not-the-retired-api-model) explains the difference.

## What a grok-imagine-image-quality request runs on now

Since November 2, 2026, xAI answers every request to `grok-imagine-image-quality` on `/v1/images/generations` and `/v1/images/edits` with `grok-imagine-image-2.0` at `quality: "low"`. Every parameter the old model accepted is still accepted, so your existing `prompt`, `n`, `aspect_ratio`, `resolution` and `response_format` keep working without edits. xAI's migration guide says the `model` field in each response reports the model that actually served the request, which gives you a way to confirm the switch from your own logs.

The same chain catches an older ID. `grok-imagine-image-pro` was retired on May 15, 2026 and redirected to `grok-imagine-image-quality`; per [xAI's May 15 retirement notice](https://docs.x.ai/developers/migration/may-15-retirement), it now follows that redirect on to 2.0 at `low`. The cheaper first-generation model, `grok-imagine-image` (1.0), is not affected on the xAI API.

What xAI has not published is how long the retired slug will keep resolving. Treat the redirect as a grace period, not a permanent alias, and move your code to the new ID while nothing is on fire.

![grok-imagine-image-pro redirects to grok-imagine-image-quality, which now runs on grok-imagine-image-2.0 at low quality, followed by three steps: swap the model ID, pin quality, confirm what ran](https://blog.laozhang.ai/posts/en/grok-imagine-image-quality-retired/img/redirect-chain-what-to-set.webp)

## What you pay after the switch: $0.04 is only 1K at low

The $0.04 figure on xAI's model page and in the plain-text version of its pricing page is the cheapest cell of a larger table: one 1K image at `low`. As of November 3, 2026, [xAI's pricing page](https://docs.x.ai/developers/pricing) bills `grok-imagine-image-2.0` per image by resolution and quality, plus $0.01 per input image on edits. Your real price depends on which of those cells your requests land in, and on whether you call xAI directly or through someone else's endpoint.

| What you send | xAI API direct | laozhang.ai (platform quote) |
| --- | --- | --- |
| Old ID, 1K, before November 2 | $0.05 | $0.045 |
| Old ID, 1K, after November 2 | $0.04 (billed as 2.0 `low`) | $0.045 (old ID still listed) |
| 2.0, `low`, 1K | $0.04 | $0.055 |
| 2.0, `low`, 2K | $0.06 | $0.055 |
| 2.0, `medium`, 1K | $0.06 | $0.055 |
| 2.0, `medium`, 2K | $0.08 | $0.055 |
| 2.0 edit, no `quality`, 1K, one source image | $0.06 + $0.01 = $0.07 | $0.055 |

*Disclosure: [laozhang.ai is our own service](https://blog.laozhang.ai/en/about#laozhang). Its prices are platform pricing, not an independent comparison.*

xAI's table also has a 1.5K tier ($0.05 at `low`, $0.07 at `medium`), although its image generation docs list only `1k` and `2k` as `resolution` values. The laozhang.ai prices come from its [Grok Imagine image API documentation](https://docs.laozhang.ai/api-capabilities/grok-imagine-image.md) as of November 3, 2026: one price for every resolution and quality, for generations and edits alike, charged per successful image with no separate input-image fee.

So the direction of your bill depends on where you call from. At 10,000 successful 1K images a month on the xAI API, the old model cost 10,000 × $0.05 = $500 and 2.0 at `low` costs 10,000 × $0.04 = $400, a $100 saving. On laozhang.ai, the same volume moves from 10,000 × $0.045 = $450 on the old ID to 10,000 × $0.055 = $550 on 2.0, $100 more. Those figures are list-price arithmetic that leaves out retries and failed calls.

![For 10,000 1K images a month, moving to grok-imagine-image-2.0 cuts the xAI API bill from $500 to $400 but raises the laozhang.ai bill from $450 to $550, with xAI's per-image 2.0 prices by setting next to laozhang.ai's flat $0.055](https://blog.laozhang.ai/posts/en/grok-imagine-image-quality-retired/img/bill-direction-xai-vs-laozhang.webp)

Your bill can still go up on the xAI API after you migrate, even though the redirect itself is cheaper, if your explicit 2.0 calls land in a dearer cell than the redirect does. Edit calls sent without `quality` are served at `medium`, so a 1K edit with one source image costs $0.07 rather than the redirect's 1K generation price, and choosing `medium` for detail costs $0.06 at 1K and $0.08 at 2K, more than the old model's $0.05 at 1K. xAI says the redirect is $0.01 cheaper than the old model at every resolution, but it has not published the old model's prices above 1K. If you are reconsidering xAI altogether, the [AI Image Generation API Comparison 2026: Pricing, Quality, Value](https://blog.laozhang.ai/en/posts/ai-image-generation-api-comparison-2026) puts other vendors' image prices side by side, and [Grok API Pricing: Model Rates, Billing Rules, and Monthly Costs](https://blog.laozhang.ai/en/posts/xai-grok-api-pricing) covers Grok's text models.

## What to set instead: grok-imagine-image-2.0 with an explicit quality

The fix xAI recommends is a one-line change to `model` plus a `quality` value you choose, so the redirect's `low` default stops deciding for you. Do it in this order, because the ID often lives in more places than the main code path:

1. Search your code, environment variables, workflow files and tool settings for `grok-imagine-image-quality` and `grok-imagine-image-pro`. ComfyUI nodes, SillyTavern image settings and gateway model maps often store the ID separately from your code.
2. Replace each one with `grok-imagine-image-2.0`, including the `.0` at the end.
3. Add `quality`: `"low"` for generations, the setting closest in cost to the old model and the one the redirect uses, `"medium"` where detail matters more than price. Set it on edit calls too, or they run at `medium`.
4. If you use the OpenAI Python SDK, pass `quality` (and `resolution`, `aspect_ratio`) inside `extra_body`; the JavaScript SDK accepts `quality` as a top-level field.
5. Confirm the change. On the xAI API, log the `model` field of each response. On a gateway, check its usage log for which model ID was billed.

A minimal generation call against the xAI API looks like this:

```bash
curl -X POST https://api.x.ai/v1/images/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $XAI_API_KEY" \
  -d '{
    "model": "grok-imagine-image-2.0",
    "prompt": "A watercolor painting of a lighthouse at dawn",
    "quality": "low"
  }'
```

The same `model` and `quality` change applies to `/v1/images/edits`. Two options are new to you after the move: 2.0 accepts up to five source images for an edit on the xAI API (laozhang.ai's endpoint accepts one to three), and it adds the `21:9` and `5:2` aspect ratios. Do not send `quality` to `grok-imagine-image` (1.0); xAI's docs say the parameter is supported only on `grok-imagine-image-2.0`.

*Disclosure: [laozhang.ai is our own service](https://blog.laozhang.ai/en/about#laozhang). Its prices are platform pricing, not an independent comparison.*

## What a call to the old ID returned on November 3, 2026

A day after the retirement, the old ID still produced images through a third-party gateway, and adding 2.0-only parameters did not cause errors. On November 3, 2026, seven requests went to laozhang.ai's endpoint `POST https://api.laozhang.ai/v1/images/generations` from one paid account in the default group, all with the prompt "A watercolor painting of a lighthouse at dawn, no text", `n: 1` and `response_format: "url"`.

*Disclosure: [laozhang.ai is our own service](https://blog.laozhang.ai/en/about#laozhang). Its prices are platform pricing, not an independent comparison.*

| Request | Status | Time | Image returned |
| --- | --- | --- | --- |
| `grok-imagine-image-quality`, no `quality` (first call) | 200 | 15.4 s | 832×1248 JPEG, about 472 KB |
| `grok-imagine-image-quality`, no `quality` (second call) | 200 | 17.3 s | 832×1248 JPEG, about 476 KB |
| `grok-imagine-image-quality` + `quality: "medium"` | 200 | 16.7 s | 832×1248 JPEG, about 503 KB |
| `grok-imagine-image-quality` + `resolution: "2k"` | 200 | 23.5 s | 1664×2496 PNG, 5.9 MB |
| `grok-imagine-image-2.0` + `quality: "low"` | 200 | 16.0 s | 832×1248 JPEG, about 474 KB |
| `grok-imagine-image-2.0` + `quality: "medium"` | 200 | 87.0 s | 1248×832 JPEG, about 430 KB |
| `grok-imagine-image-2.0`, no `quality` | 200 | 14.6 s | 832×1248 JPEG, about 461 KB |

Every response body held only `created`, `data[].url` and a `usage` block reporting 1,000 tokens. There was no `model` field, so these calls cannot show whether the old ID was served by 2.0 at `low`; laozhang.ai's own documentation lists it as a separate model at its old price. xAI's statement about the `model` field was not tested, because no call went to the xAI API directly. The results come from one gateway, one account, one prompt and one day, with a single sample per setting; the 87-second `medium` call in particular is one data point, not a speed measurement. Billing was not read back per call.

## Will images look worse on grok-imagine-image-2.0 at low?

xAI has not published a same-prompt comparison of the retired model against 2.0 at `low`, so whether redirected output looks worse is unknown as of November 3, 2026. xAI's position is that `low` is the closest cost match to the old model and ranks #2 on the Arena image-edit leaderboard and #3 on text-to-image; those ranks are xAI's own statement.

In the November 3 calls above, the images from the old ID and from 2.0 at `low` looked alike in style and composition when viewed side by side, but that was one prompt judged by eye, without a blind comparison. If your output looks softer after the switch and detail matters for that job, set `quality: "medium"` on those requests and budget $0.06 per 1K image on the xAI API.

## Grok app Quality Mode is not the retired API model

The Grok app's Quality Mode and the `grok-imagine-image-quality` API model only share a word. On August 8, 2026, TestingCatalog reported, quoting xAI's @grok account, that [Imagine Image 2.0 became Grok's Quality Mode](https://www.testingcatalog.com/xai-launches-imagine-image-2-0-in-grok-quality-mode/) on the web and in the iOS and Android apps. xAI's November 2 notice covers only the API slug, and nothing published says Quality Mode changed that day; that it was not retired is an inference from that silence, not an xAI statement. Questions about how many images the app lets you make and when that resets belong to [Grok Free Limit: What xAI Publishes, How It Resets, What to Do](https://blog.laozhang.ai/en/posts/grok-free-limit).

## Sources

External pages this guide links to, in the order they appear. Last updated 2026-11-03.

- [xAI's grok-imagine-image-quality migration guide](https://docs.x.ai/developers/migration/imagine-image-quality-nov-2) (docs.x.ai)
- [xAI's May 15 retirement notice](https://docs.x.ai/developers/migration/may-15-retirement) (docs.x.ai)
- [xAI's pricing page](https://docs.x.ai/developers/pricing) (docs.x.ai)
- [Imagine Image 2.0 became Grok's Quality Mode](https://www.testingcatalog.com/xai-launches-imagine-image-2-0-in-grok-quality-mode/) (testingcatalog.com)