Image

GPT Image 2 API

Reasoning-first images, with text that actually reads.

Generate and edit images with OpenAI's GPT Image 2 through one unified API. A single REST endpoint, async jobs, and webhooks, with no separate OpenAI account to wire up.

Get API key Read the docs

8
Images per prompt: 2K
Max via API: ~30s
Avg. generation: 2 credits
From

GPT Image 2 photoreal ATLAS magazine cover on a rainy European newsstand, bilingual English and Japanese cover lines

Playground

Try GPT Image 2 right here.

model: gpt-image-2

Prompt

Aspect ratio

Quality

Background

Number of images

Media inputs

Reference image

Advanced settings

Output format Moderation

Estimated cost 2–16 credits · $0.02–$0.16 Generate in Studio

Output Example

GPT Image 2 bilingual ATLAS FEST 2026 poster with English and Japanese type, gold rules, and a lighthouse — A bilingual event poster for ATLAS FEST 2026, literary festival. Headline ATLAS FEST in huge condensed black type, Japanese line 文学祭り 2026 beneath it perfectly legible. Subcopy: Palace Theatre, Sept 21–23, dusk indigo gradient, gold accent rules, a small lighthouse icon, date and venue block at the bottom, print-ready, no extra logos, no misspellings.

Capabilities

What's new in GPT Image 2.

OpenAI's latest flagship image model

It powers ChatGPT Images 2.0 and replaces both DALL-E 3 and the interim GPT Image 1.5, available in ChatGPT and via the API as gpt-image-2.

Reasoning built in

It is the first OpenAI image model with a "thinking" mode, so before generating it can plan the layout, search the web for references, and self-check its output.

Up to 8 images from one prompt

It can produce a coherent set of images in a single shot, keeping characters and objects consistent across all of them, which suits storyboards and brand campaigns.

State-of-the-art text rendering

It handles dense and multilingual text cleanly across multiple scripts, making it dependable for posters, packaging, menus, and UI labels.

Precise instruction following

OpenAI describes it as a step change in following detailed prompts and in placing and relating objects accurately within a scene.

Higher resolution and flexible editing

It supports up to 2K resolution via the API with a range of aspect ratios, plus multi-turn editing that preserves identity and context across changes.

Showcase

See what GPT Image 2 can create.

Real outputs generated with GPT Image 2 on Apiframe, each with the prompt behind it.

Try GPT Image 2

GPT Image 2 four-panel comic of Ink the office cat taking over the ATLAS layout table — A four-panel comic strip with readable handwritten-style dialogue. A grey office cat named Ink refuses to get off the ATLAS magazine layout table. Panel 1: editor asks politely. Panel 2: cat sits on the proofs. Panel 3: cat types gibberish on the keyboard. Panel 4: the printed cover has a paw print and the cat looks proud. Clean ink lines, cream paper, all lettering perfectly legible, no misspellings.

GPT Image 2 character sheet of the ATLAS fox mascot in eight consistent poses — A character sheet of the same ATLAS fox mascot in eight consistent poses: waving, running, sitting, jumping, reading a magazine, holding a coffee, sleeping on proofs, pointing at a headline. Flat graphic design, rust and navy palette, white background, small pose labels under each figure in clean sans type, identical character in every pose, no misspellings.

GPT Image 2 infographic titled How a Magazine Is Made, five labeled production stages — An infographic titled HOW A MAGAZINE IS MADE, five labeled stages: Commission, Edit, Design, Print, Ship. Clean editorial layout, numbered steps, tiny caption text that is fully readable, rust and navy on cream paper, ATLAS wordmark small at the top, no decorative clutter, no misspellings.

GPT Image 2 desktop landing-page mockup for atlas.press with Essays, Photography, and Dispatch cards — A desktop landing-page mockup for atlas.press, a literary magazine site. Hero header ATLAS, subhead Stories that travel, three feature cards titled Essays, Photography, Dispatch, a Subscribe button, rust and navy UI, browser chrome around a 1440px layout, sharp UI type, no lorem ipsum, no misspellings.

GPT Image 2 print-ready dust jacket for The Last Lighthouse by Ella Marlow, published by ATLAS PRESS — A hardcover book dust jacket for THE LAST LIGHTHOUSE, published by ATLAS PRESS. Title in large serif across the upper third, author name ELLA MARLOW, a painterly lighthouse in fog, spine visible at left with ATLAS PRESS, back-cover blurb in small but fully readable serif, print-production still on a wooden table, no misspellings.

Pricing

Simple per-generation pricing.

2–16 credits per generation ≈ $0.02–$0.16

The rate depends on the resolution and options you pick.

You only pay for successful generations. Failed jobs are refunded automatically.
One credit balance across every model on Apiframe, no per-model plans.
Start with free credits. No subscription required.
Volume discounts on larger plans.

See plans and credit packs API pricing GPT Image 2 credit cost and rates

Specs

At a glance.

Provider OpenAI

Model ID gpt-image-2

Aspect ratios 1:1, 3:2, 2:3

Image input Supported

Avg. time ~55s

Comparison

GPT Image 2 vs Nano Banana 2 vs Flux 2 Pro

GPT Image 2 against the Google and Flux flagships. All on the same key.

Feature	GPT Image 2	Nano Banana 2	Flux 2 Pro
Max resolution	2K via API	4K	4MP
Avg. generation	~30s	~12s	~15s
Signature strength	Reasoning + batch of 8 + text accuracy	Speed/quality sweet spot + search grounding	Photoreal volume at a balanced price
Price per image	2–17 credits (≈$0.02–$0.17)	8–19 credits (≈$0.08–$0.19)	3–9 credits (≈$0.03–$0.09)
Best for	Instruction-heavy, text-accurate work	Everyday generation and editing	High-volume production at a balanced price

Specs reflect what each model supports on Apiframe today, from the live catalog.

Quickstart

How to call the GPT Image 2 API.

Send one POST /v2/images/generate request with your API key. The call returns a jobId you can poll, or pass a webhook_url and the result is pushed to you the moment it's ready.

Swap gpt-image-2 for any other model and nothing else changes.

Get API key Read the docs

generate.js

const response = await fetch("https://api.apiframe.ai/v2/images/generate", {
  method: "POST",
  headers: {
    "X-API-Key": "afk_your_api_key_here",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    "prompt": "a sleek silver sports car on a coastal highway at sunset, hyper-realistic",
    "model": "gpt-image-2",
    "gptImage2Params": {
      "input_images": "https://example.com/input.jpg",
      "quality": "auto",
      "background": "auto",
      "number_of_images": 1
    }
  }),
});
const { jobId } = await response.json();
console.log(jobId);

import requests

response = requests.post(
    "https://api.apiframe.ai/v2/images/generate",
    headers={
        "X-API-Key": "afk_your_api_key_here",
        "Content-Type": "application/json",
    },
    json={
        "prompt": "a sleek silver sports car on a coastal highway at sunset, hyper-realistic",
        "model": "gpt-image-2",
        "gptImage2Params": {
            "input_images": "https://example.com/input.jpg",
            "quality": "auto",
            "background": "auto",
            "number_of_images": 1
        }
    },
)
print(response.json())  # { "jobId": "...", "status": "QUEUED" }

curl -X POST https://api.apiframe.ai/v2/images/generate \
  -H "X-API-Key: afk_your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
        "prompt": "a sleek silver sports car on a coastal highway at sunset, hyper-realistic",
        "model": "gpt-image-2",
        "gptImage2Params": {
            "input_images": "https://example.com/input.jpg",
            "quality": "auto",
            "background": "auto",
            "number_of_images": 1
        }
    }'

let job;
do {
  await new Promise((r) => setTimeout(r, 2000));
  job = await fetch("https://api.apiframe.ai/v2/jobs/JOB_ID", {
    headers: { "X-API-Key": "afk_your_api_key_here" },
  }).then((r) => r.json());
} while (job.status !== "COMPLETED" && job.status !== "FAILED");
console.log(job.result);

import requests, time

while True:
    job = requests.get(
        "https://api.apiframe.ai/v2/jobs/JOB_ID",
        headers={"X-API-Key": "afk_your_api_key_here"},
    ).json()
    if job["status"] in ("COMPLETED", "FAILED"):
        break
    time.sleep(2)
print(job["result"])

curl https://api.apiframe.ai/v2/jobs/JOB_ID \
  -H "X-API-Key: afk_your_api_key_here"

Input schema

Every field GPT Image 2 accepts, with types and defaults. The full reference lives in the docs.

Field	Type	Description
prompt required	string	Text description of what to generate.
model required	string	The model identifier for this endpoint. Default:"gpt-image-2" · "gpt-image-2"
gptImage2Params.input_images	string (URL)	Reference image (URL)
gptImage2Params.quality	string	Quality Default:"auto" · "auto", "low", "medium", "high"
gptImage2Params.background	string	Background Default:"auto" · "auto", "opaque"
gptImage2Params.number_of_images	number	Number of images Default:1 · min 1, max 10, step 1
gptImage2Params.output_format	string	Output format Default:"webp" · "webp", "png", "jpeg"
gptImage2Params.moderation	string	How strictly to filter content. Use "low" for less restrictive moderation. Default:"auto" · "auto", "low"

GPT Image versions

Every GPT Image image model on Apiframe. Switch with one parameter.

Use cases

What teams build with GPT Image 2.

Instruction-heavy briefs

Built-in thinking plans layout, object placement, and self-checks before rendering, so long multi-part prompts land as specified instead of approximately.

Posters, packaging, UI type

State-of-the-art dense and multilingual text rendering makes it the pick for covers, menus, comics, and interface mockups where copy has to read.

Consistent sets in one shot

Up to eight images from a single prompt, characters and objects held across the set, storyboards and campaign variants without a separate consistency pass.

Multi-turn editing

Identity and scene hold through sequential edits, add a coat, restage a product, keep the face, so production can iterate without starting over.

Reviews

9,800+ developers. One API.

Real Apiframe reviews from Trustpilot and G2.

Speeds up our ad creative iteration

The API integrates directly into my existing pipeline, so I can generate and iterate on image variations for Meta ads without jumping between tools. We can go from concept to a testable ad in the same day instead of spreading it across a week.

Ethan M. July 2026 G2

Fast, reliable, easy to integrate

Image generation requests are processed quickly, and the results are consistently high quality. Customer support has also been responsive whenever I had questions. It has saved me a lot of development time.

Nana Arianto July 2026 Trustpilot

Seamless integrations, effortless setup

Generating ad creatives for Facebook and Google ads is more streamlined now since we use APIFRAME to manage the workflow, which used to require a lot of time with official APIs. The initial setup was very easy, just a matter of using curl.

Bean A. July 2026 G2

GPT Image 2 questions

What teams ask before shipping with GPT Image 2. Everything else lives in the docs.

What is GPT Image 2?

OpenAI's newest image generation and editing model, released in April 2026 as the engine behind ChatGPT Images 2.0, replacing DALL-E 3 and GPT Image 1.5.

What is "thinking" mode?

A reasoning step where the model plans the layout, can search the web for references, and reviews its own output before finishing, and it enables generating up to eight images from one prompt.

How is it different from GPT Image 1.5?

It is the newer model, adding built-in reasoning, stronger multilingual text, consistent multi-image sets, and higher overall quality.

Can it render text in multiple languages?

Yes. Multilingual text rendering across several scripts is one of its standout strengths.

Is it free?

Standard mode is available to ChatGPT users at no extra cost, while the thinking features are reserved for paid tiers, and API access is paid on a token basis.

Where can you access it?

Through Apiframe, as well as ChatGPT and the OpenAI API.