GPT Image 2
⧉GPT Image 2 is built for commercial creative work: article covers, product shots, posters, infographics and illustrations. Supports both text-to-image and image-to-image, 1K / 2K / 4K resolutions and multiple aspect ratios, through the OpenAI-compatible images API.
What is the GPT Image 2 API and what does it cost?
GPT Image 2 is OpenAI's image generation and editing model: text-to-image, and image-to-image from up to 8 reference images. On NezhaGate it costs $0.005 per image at 1K ($0.025 at 2K, $0.04 at 4K) (OpenAI list price for a 1024x1024 image: $0.006 low, $0.053 medium or $0.211 high quality, plus input tokens). Base URL: https://nezhagate.com/v1 (OpenAI-compatible).
Measured performance
Real image jobs sent through the public NezhaGate API. Test window 2026-10-20 18:17 – 2026-10-20 18:21 UTC; sample size 8.
How this was measured
/v1/images/generations at 1K, at most four in flight at once. Time is measured from submitting the job to the moment polling first sees status: "succeeded", so it includes any queueing; it is only as precise as the polling interval (3 s). With a sample this small, read the numbers as indicative.
Measurements are repeated every week and this page updates itself; the date above is the latest run.
Cost example
NezhaGate figures use the live price on this page; vendor figures use the vendor's published list price (sources below).
1,000 square images at 1K
1,000 images at 1024x1024. OpenAI prices GPT Image 2 by quality level and bills input tokens on top; NezhaGate charges one flat price per image whatever the quality: images are drawn at medium by default, and at 1K you can choose low or high at the same price.
| Where | Cost | Difference |
|---|---|---|
| NezhaGate | $5.00 | — |
| OpenAI, low quality (+ input tokens) | $6.00 | NezhaGate 17% cheaper |
| OpenAI, medium quality (+ input tokens) | $53.00 | NezhaGate 91% cheaper |
| OpenAI, high quality (+ input tokens) | $211.00 | NezhaGate 98% cheaper |
Vendor list prices checked 2026-09-25: OpenAI image generation guide (costs), OpenAI API pricing.
Parameters & capabilities
"Vendor spec" is what the model's maker publishes. "Via NezhaGate" is what our own requests through the gateway showed (last checked 2026-10-20).
| Item | Vendor spec | Via NezhaGate |
|---|---|---|
| Model id | — | gpt-image-2 |
| Endpoints | — | POST /v1/images/generations (text-to-image) and POST /v1/images/edits (image-to-image); both return a job you poll at GET /v1/images/jobs/{id} |
| Price | By quality: $0.006 / $0.053 / $0.211 per 1024x1024 image, plus input tokens OpenAI image generation guide (costs) | $0.005 at 1K · $0.025 at 2K · $0.04 at 4K, flat per image |
| Sizes | — | An aspect ratio (1:1, 3:4, 2:3, 9:16, 4:3, 3:2, 16:9) or pixels (1024x1536); resolution 1K / 2K / 4K sets the long side to 1024 / 2048 / 4096 px |
| Quality setting | low / medium / high OpenAI image generation guide (costs) | low, medium (default) or high at 1K, same price; 2K and 4K always medium |
| Reference images | — | Up to 8 per request, 16 MB each: public URL, data: URI or base64 |
| Output | — | PNG at a stable img.nezhagate.com URL; add response_format: "b64_json" to also receive base64 |
| Response mode | — | Asynchronous: HTTP 202 with a job id, then poll |
| Failed jobs | — | Refunded in full |
Tool setup
Base URL https://nezhagate.com/v1, your NezhaGate API key, and the model id gpt-image-2. Request and response samples (cURL, Python, Node.js) are in the GPT Image 2 API docs.
- Add an HTTP Request node: method
POST, URLhttps://nezhagate.com/v1/images/generations, headerAuthorization: Bearer <your key>, JSON body withmodel=gpt-image-2and yourprompt. - Add a Wait node (3 seconds), then a second HTTP Request node:
GET https://nezhagate.com/v1/images/jobs/{{ $json.id }}with the same header. - Add an IF node on
{{ $json.status }}:succeededcontinues with{{ $json.data[0].url }}; anything else loops back to the Wait node (stop onfailed).
Differences from the official API & known limits
- Job-based instead of synchronous: OpenAI's Images API returns the image in the response; NezhaGate returns HTTP 202 with a job id and you poll
GET /v1/images/jobs/{id}untilstatusissucceeded. Code written for the synchronous response needs the short polling loop shown above. qualitydoes not change the price (low/medium/highat 1K, medium at 2K and 4K); the price depends only onresolution(1K / 2K / 4K), and there is no separate charge for input tokens or reference images.- Any aspect ratio is accepted; the result is centre-cropped and resized to exactly the size you asked for and saved as PNG.
- An edit request takes at most 8 reference images; any beyond 8 are dropped without an error.
- Image URLs are hosted on
img.nezhagate.com; the job record itself can be polled for 3 days after it finishes. - NezhaGate also serves OpenAI's newer GPT Image 2.5 Flare and GPT Image 2.5 Sunburst on the same API.
GPT Image 2 API: frequently asked questions
How long does GPT Image 2 take to generate an image on NezhaGate?
Why do I get a job id instead of an image?
GET /v1/images/jobs/{id} every few seconds until status is succeeded, then read data[0].url.How much does GPT Image 2 cost?
Can I choose the quality level?
quality as low, medium (the default) or high, at the same price. 2K and 4K are always drawn at medium. The price is set by resolution (1K, 2K or 4K).What sizes can I request?
1:1, 3:4 or 16:9, or a pixel size such as 1024x1536. With resolution 1K the long side is 1024 px, with 2K 2048 px and with 4K 4096 px.Real images from the test run
Generated through the API on 2026-10-20 with the prompts shown; not edited or cherry-picked beyond one result per prompt.
A minimalist poster for a jazz night called "BLUE HOUR", bold sans-serif title, deep navy and brass colours, one saxophone silhouette, lots of negative space
size 3:4 · 1K · 768×1024 px · ready in 26.3 s
Studio product photo of a matte ceramic coffee mug on a light oak table, soft window light from the left, shallow depth of field
size 1:1 · 1K · 1024×1024 px · ready in 29.6 s
Wide cinematic landscape of terraced rice fields at sunrise, low mist in the valleys, warm golden light, a small farmhouse in the middle distance
size 16:9 · 1K · 1024×576 px · ready in 25.5 s
Flat vector infographic titled "HOW A REQUEST TRAVELS" showing four labelled steps: Client, CDN, Gateway, Model, connected by arrows, clean pastel palette
size 4:3 · 1K · 1024×768 px · ready in 30.9 s
Live Test · Playground
Try out GPT Image 2 right here (available after login).
Input
ImageOutput
imageAbout GPT Image 2
GPT Image 2 is OpenAI's high-quality image generation model for both text-to-image and image-to-image workflows, producing finished covers, posters, and illustrations. On NezhaGate it is served through an OpenAI-compatible endpoint: point your base_url at NezhaGate and set the model to "gpt-image-2" to call it with your existing OpenAI SDK. Billing is pay-as-you-go, and failed requests are not charged.
Use cases
Turn a short text brief into event posters, article covers, or product key visuals, and generate several variations quickly without designing from scratch.
Generate scene shots, background swaps, or styled product imagery to support detail pages and ad placements.
Create illustrations and diagrams that match your content, making technical docs, blog posts, and decks easier to follow.
Provide a reference image with a prompt to redraw composition, style, or details, producing consistent variants fast.
Batch-generate visuals for different platform sizes and themes while keeping a consistent brand look.
How to choose
GPT Image 2 is the catalogue's dedicated image model, so reach for it whenever the output is a picture. When you need text chat, reasoning, or code, choose a chat model instead, such as the flagship GPT-5.5, Claude Sonnet 4.6, or Gemini 3.1 Pro; in workflows that need both copy and visuals, pair one of those with gpt-image-2.
Related Models
Explore other models you can integrate.
GPT Image 2.5 Flare
Next-gen image model - clean and smooth
Nano Banana Pro
Flagship text-to-image / image-to-image model
Nano Banana 2
High-quality text-to-image / image-to-image model
GPT Image 2.5 Sunburst
Next-gen image model - richer texture
Grok Imagine Image
xAI's image model · renders in seconds
