grok-4.2-imagegrok-4.2-image is the budget tier for Grok image generation on apimodels.app: $0.0075 per image, a quarter of what the standard Grok Imagine Image tier costs, running on the same model behind the same endpoint. Send a prompt for text-to-image, or add image_url and the same call edits instead — 19 seconds at the median in production. Aspect ratio is controllable across ten values (1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 2:1, 1:2, auto), verified one ratio at a time against the live channel rather than copied from a spec sheet. Resolution takes 1k or 2k and defaults to 2k, at the same flat price either way — a 2K frame (2048x2048 square, 2816x1584 at 16:9) costs the same $0.0075 as a 1K one. One API key, callable from mainland China without an xAI account, and failures are never charged. Three things it does not do: the four ultra-wide ratios (19.5:9, 9:19.5, 20:9, 9:20 — use grok-imagine-image), mask-based inpainting (that belonged to a retired channel, not to the model), and multi-image fusion. When you need higher fidelity, grok-imagine-image-pro is a one-field switch away.
$0.0075 per image against $0.03 on the standard Grok Imagine Image tier — same model, same endpoint. Switching tiers is a one-field change, so drafts can run cheap and only the final render needs the expensive tier
Create-then-poll by default; pass an OpenAI-style size (e.g. 1024x1024) with no callback_url and the single request blocks until the image is ready instead. Measured in production: 19s median, 22s at p90, tail to 28s — set the client timeout above 60s
Prompt only is text-to-image; add image_url (or image_base64) and the same call edits instead — no endpoint switch, no separate model name. Editing holds the input: framing, camera angle, background and lighting survive, so a recolour stays a recolour
Pass resolution 1k or 2k — the price is flat either way, so a 2K frame costs the same $0.0075 as a 1K one. Measured output: 2048x2048 for a square 2K request, 2816x1584 at 16:9. Omit the field and you get 2k by default. For comparison, grok-imagine-image-pro charges $0.12 for its 2K tier
1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 2:1, 1:2 and auto — square, portrait, landscape and the 2:1 shape social previews want. Measured against the live channel, not copied from a spec sheet: the four ultra-wide ratios (19.5:9, 9:19.5, 20:9, 9:20) are rejected on this tier, and grok-imagine-image is where you go for those
High-volume drafting and iteration, social and ad creative in odd ratios, thumbnail and banner batches, and any pipeline where per-image cost decides whether the feature ships
Mask-based inpainting (not offered — pass the original and describe the change in words), multi-image fusion (single reference only), or work needing the photoreal ceiling — use grok-imagine-image-pro for that
Both tiers bill the same flat $0.0075 — 2K costs no extra (2048x2048 square, 2816x1584 at 16:9). 1K is about twice as fast.
Generated image will appear here
Enter a prompt and click Generate
Grok 4.2 Image は xAI の画像生成 API です。grok-4.2-image is the budget tier for Grok image generation on apimodels.app: $0.0075 per image, a quarter of what the standard Grok Imagine Image tier costs, running on the same model behind the same endpoint. Send a prompt for text-to-image, or add image_url and the same call edits instead — 19 seconds at the median in production. Aspect ratio is controllable across ten values (1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 2:1, 1:2, auto), verified one ratio at a time against the live channel rather than copied from a spec sheet. Resolution takes 1k or 2k and defaults to 2k, at the same flat price either way — a 2K frame (2048x2048 square, 2816x1584 at 16:9) costs the same $0.0075 as a 1K one. One API key, callable from mainland China without an xAI account, and failures are never charged. Three things it does not do: the four ultra-wide ratios (19.5:9, 9:19.5, 20:9, 9:20 — use grok-imagine-image), mask-based inpainting (that belonged to a retired channel, not to the model), and multi-image fusion. When you need higher fidelity, grok-imagine-image-pro is a one-field switch away. APIMODELS のプラットフォーム経由なら、統一 API と明朗な従量課金でこのモデルを呼び出せます。 現在の料金: per image: $0.0075。
At $0.0075 a frame you can explore twenty directions for less than one image on the Pro tier. When a draft is right, change the model field to grok-imagine-image-pro and re-render that single prompt at full fidelity — same endpoint, same parameters.
One product photo into image_url, then one call per colour, material or season. Framing, camera angle, background and lighting hold, so the set looks shot together rather than generated separately.
A/B thumbnail variants, blog headers, category banners: 16:9 and 2:1 with in-image text that survives at small sizes. A hundred thumbnails costs $0.75.
9:16 and 3:4 for stories, reels and pins, with room left in frame for the caption. Ad creative where you need thirty variants to find two that work.
Seed a catalogue, a design system or a staging database with plausible imagery instead of grey boxes. Cheap enough to regenerate the whole set when the brief changes.
Ultra-wide wallpapers (19.5:9, 20:9) need grok-imagine-image. Combining several references into one scene needs gpt-image-2 (16) or grok-imagine-image-pro (3). Masked local repainting is not offered on any Grok tier here.
Grok 4.2 Image は APIMODELS 経由で per image: $0.0075 で利用できます。課金は従量制で、生成した分だけの支払いです。
APIMODELS に登録して API キーを取得し、統一エンドポイントを呼ぶだけです。cURL / Python / Node.js のサンプルを含む詳しいドキュメントを用意しています。
APIMODELS は同じ Grok 4.2 Image を集約プラットフォーム経由で提供します。統一された API インターフェースなので、プロバイダごとにアカウントを作る必要はありません。キー 1 本ですべてのモデルに届きます。
It is the budget tier for Grok image generation on apimodels.app: $0.0075 per image, a quarter of the standard Grok Imagine Image price ($0.03), running on the same model behind the same endpoint. Text-to-image and single-reference editing, 19 seconds at the median in production. One clarification worth stating plainly: xAI publishes no model called grok-4.2-image (its image models are grok-imagine-image, -2.0 and -quality). The name is a callable model on this platform, and the budget tier lives on it.
A flat $0.0075 per image — independent of aspect ratio and prompt length, and the same price when you pass a reference image to edit. Failures are never charged, and there is no minimum top-up or subscription. For comparison: the standard tier grok-imagine-image is $0.03 per image, and grok-imagine-image-pro is $0.08 at 1K / $0.12 at 2K.
Same upstream model; what differs is price and the capability ceiling. For volume work, drafting, and batches of thumbnails or banners, grok-4.2-image at $0.0075 goes furthest. Take grok-imagine-image ($0.03) when you need the wider ratio set — including ultra-wide 19.5:9 and 20:9 — or general-purpose reliability. Use grok-imagine-image-pro ($0.08 / $0.12, up to 3 references) for photoreal detail and 2K. All three share one endpoint and the same parameters, so switching is a one-field change — iterate on the cheap tier, render the final frame on Pro.
POST https://api.apimodels.app/v1/images/generations with Authorization: Bearer YOUR_API_KEY and body {"model":"grok-4.2-image","prompt":"..."}. By default it answers with a taskId and state:"pending"; GET the same path with ?task_id=xxx to collect the result. To get the image back from a single call, add an OpenAI-style size (e.g. "1024x1024") and no callback_url — the request then blocks until the image is ready and returns the OpenAI shape {created, data:[...]}. Add response_format:"url" if you want a link rather than base64.
Put the image address in **image_url** (or pass base64 in image_base64) and describe the change in the prompt. Same model name, same endpoint. One trap worth naming: the field is image_url — spell it image and it is silently dropped, so you get a plain text-to-image result and still pay for it. Fidelity is good: asked to hold composition, camera angle, background and lighting and only change a vessel from matte black to terracotta plus three coffee beans beside it, everything outside that instruction survived.
Ten values: 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 2:1, 1:2 and auto. That list came from calling the live channel one ratio at a time rather than from a spec sheet — the four ultra-wide ratios (19.5:9, 9:19.5, 20:9, 9:20) are rejected on this tier, and grok-imagine-image is the one to use when you need them. Omit aspect_ratio to take the upstream default; when editing, framing follows the reference image.
Yes, and no extra charge. Pass resolution as 1k or 2k — **omit it and you get 2k by default** — and either tier bills the same flat $0.0075. Measured output: 2048x2048 (4.19MP) for a square 2K request and 2816x1584 at 16:9; 1K squares come back 1024x1024. For contrast, grok-imagine-image-pro charges $0.12 for its 2K tier, so the same resolution is 16x cheaper here. Latency is in the same ballpark for both tiers (production medians: 13s on the default tier, 7s at 1K) with a tail past 100s either way, so keep the client timeout above 60 seconds. One trap worth knowing: if you pass an OpenAI-style size without an explicit resolution, the tier is derived from that size — send 1024x1024 and you land on 1K. Set resolution explicitly when you want 2K.
Neither, to be blunt. Mask inpainting belonged to a channel that has since been retired, not to the model — to change a region, pass the original in image_url and describe the change in the prompt. Multi-image fusion is out too: one reference image only. If you need several, use gpt-image-2 (up to 16) or grok-imagine-image-pro (3).
Measured in production: 19 seconds at the median, 22 at p90, with a tail out to 28 and a floor around 6. **Set your client timeout above 60 seconds** rather than leaving it at a 10- or 30-second default. On the default create-then-poll path the create call returns immediately and only the generation is slow. Real problems surface as an error: reword the prompt on a content-safety refusal, and retry on an upstream-busy signal. Neither is charged.
Seven days, then the files are deleted automatically. Download and re-host anything you need to keep — do not treat our URLs as permanent image hosting in your own database or pages.
APIMODELS では Grok 4.2 Image が 60 以上のモデルと同じ API キー・同じ残高の上に並んでいるので、選択は「相性」の問題であって「囲い込み」の問題ではありません。Text to Image、Image Editing、Native 1K / 2K、Cheapest Grok Tier に対応しており、他の画像生成モデルと価格・機能を並べて評価できます。乗り換えはモデル名の文字列を 1 つ書き換えるだけ。新しいアカウントも追加の実装も要りません。画像生成の選択肢と最新価格は apimodels.app/models で確認できます。
Grok 4.2 Image は次に対応しています: Text to Image、Image Editing、Native 1K / 2K、Cheapest Grok Tier。パラメータの全一覧と呼び出し例は APIMODELS のドキュメントをご覧ください。
はい。APIMODELS は Grok 4.2 Image を単一の統一 API とキー 1 本で提供します。プロバイダごとのアカウントも、各社の地域ごとのネットワーク経路を自分で面倒みる必要もありません。
Stripe（Visa、Mastercard などの国際カード）と Alipay に対応しています。支払い後、残高はすぐ反映されます。