GPT Image 2 API: settings, cost and integration

Read in Russian

Published: Oct 24, 2026, 3:47 AM UTC · Updated: Oct 24, 2026, 9:25 AM UTC

Exact ID: gpt-image-2. Check access for your key using GET /v1/models; a published model does not guarantee access for every account.

Prices and billing units

API429 catalog generated Oct 25, 2026, 7:04 AM UTC. The table shows direct request components; other general catalog metrics are not added here. Display prices are rounded to six decimal places; calculations use exact catalog rates. The calculator below follows the direct Images request contract. Sources checked: Oct 24, 2026.

OpenAI Standard: $5 per 1M text input tokens, $8 per 1M image input tokens and $30 per 1M output image tokens. The calculator uses published output image estimates for explicit common sizes; this is not a fixed bill. Pricing sources are listed below. Batch, cache and special modes are kept separate from the base comparison.

Image generation budget

Specifications and compatibility boundaries

• According to OpenAI documentation reviewed on 2026-10-24, OpenAI: text and image input, image output. The reviewed exact model page does not publish a numerical context or text output limit.
• According to OpenAI documentation reviewed on 2026-10-24, native dimensions: sides divisible by 16, maximum side 3840, aspect ratio at most 3:1, and 655,360 to 8,294,400 pixels. Modes above 2560×1440 are described as experimental.
• OpenAI’s model documentation reviewed on 2026-10-24 lists low/medium/high/auto, PNG/JPEG/WebP and preview transparency. API429 client guidance reviewed on 2026-10-24 recommends opaque/auto; native preview status does not promise API429 transparency.

Specification: OpenAI, checked Oct 24, 2026. These are not verified effective API429 limits. Check native manufacturer parameters and capabilities separately for the client method.

Published capabilities for this exact ID: chat image_generation . A missing field is not evidence of a prohibition, but it does not establish compatibility.

Public model observations

Over the last 24 hours: mean successful response time — 34 s; service error rate — 1.5%. Data as of Oct 25, 2026, 7:02 AM UTC.

Source: public 24-hour statistics. The mean is neither a median nor a response time guarantee; it is not a quality assessment of the image.

First request: curl, Python, JavaScript

Set API429_API_KEY on the server and check the exact ID using GET https://gateway.api429.com/v1/models using the same Bearer key. Generation incurs a charge; there are no automatic retries. An uncertain timeout does not prove there was no charge.

The examples request one result and b64_json. curl saves JSON to image-response.json; Python and JavaScript decode the file after checking MIME by signature. When choosing url, first distinguish a hosted URL from a data URI. Check warnings, partial results, api429.outputs and api429.parameter_mismatches, if present.

curl

curl --fail-with-body https://gateway.api429.com/v1/images/generations \
  -H "Authorization: Bearer $API429_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-image-2","prompt":"A ceramic cup on a light wooden table, soft daylight, editorial product photograph, no text.","n":1,"size":"1024x1024","response_format":"b64_json","quality":"low"}' \
  --output image-response.json

Python 3

import base64
import json
import os
import urllib.request
from pathlib import Path

key = os.environ["API429_API_KEY"]
base = "https://gateway.api429.com/v1"
headers = {"Authorization": "Bearer " + key, "Content-Type": "application/json"}
model = "gpt-image-2"
with urllib.request.urlopen(urllib.request.Request(base + "/models", headers=headers), timeout=30) as response:
    available = {item["id"] for item in json.load(response)["data"]}
if model not in available:
    raise ValueError("Model unavailable to this key")
payload = json.loads("{\"model\":\"gpt-image-2\",\"prompt\":\"A ceramic cup on a light wooden table, soft daylight, editorial product photograph, no text.\",\"n\":1,\"size\":\"1024x1024\",\"response_format\":\"b64_json\",\"quality\":\"low\"}")
request = urllib.request.Request(base + "/images/generations",
    data=json.dumps(payload).encode("utf-8"), headers=headers, method="POST")
with urllib.request.urlopen(request, timeout=180) as response:
    result = json.load(response)
if result.get("warnings") or result.get("api429", {}).get("parameter_mismatches"):
    print("Inspect output warnings:", result.get("warnings"), result.get("api429"))
items = result.get("data", [])
if len(items) != 1 or not items[0].get("b64_json"):
    raise ValueError("Missing image or partial result; inspect before retrying")
raw = base64.b64decode(items[0]["b64_json"], validate=True)
if raw.startswith(b"\x89PNG\r\n\x1a\n"):
    extension = "png"
elif raw.startswith(b"\xff\xd8\xff"):
    extension = "jpg"
elif raw.startswith(b"RIFF") and raw[8:12] == b"WEBP":
    extension = "webp"
else:
    raise ValueError("Unknown image MIME; inspect before saving")
Path("image." + extension).write_bytes(raw)
print(result.get("api429", {}).get("outputs"))

JavaScript: Node.js 18+

import { writeFile } from "node:fs/promises";
const key = process.env.API429_API_KEY;
if (!key) throw new Error("Set API429_API_KEY");
const base = "https://gateway.api429.com/v1";
const headers = { Authorization: "Bearer " + key, "Content-Type": "application/json" };
const model = "gpt-image-2";
const available = await fetch(base + "/models", { headers, signal: AbortSignal.timeout(30000) });
if (!available.ok) throw new Error("Models HTTP " + available.status);
if (!(await available.json()).data.some(item => item.id === model)) throw new Error("Model unavailable to this key");
const payload = {"model":"gpt-image-2","prompt":"A ceramic cup on a light wooden table, soft daylight, editorial product photograph, no text.","n":1,"size":"1024x1024","response_format":"b64_json","quality":"low"};
const response = await fetch(base + "/images/generations", {
 method: "POST", headers, body: JSON.stringify(payload), signal: AbortSignal.timeout(180000)
});
if (!response.ok) throw new Error("Generation HTTP " + response.status);
const result = await response.json();
if (result.warnings || result.api429?.parameter_mismatches) console.warn("Inspect warnings", result.warnings, result.api429);
if (result.data?.length !== 1 || !result.data[0].b64_json) throw new Error("Missing image or partial result; inspect before retrying");
const bytes = Buffer.from(result.data[0].b64_json, "base64");
const hex = bytes.subarray(0, 8).toString("hex");
const extension = hex === "89504e470d0a1a0a" ? "png" : hex.startsWith("ffd8ff") ? "jpg" : bytes.toString("ascii", 0, 4) === "RIFF" && bytes.toString("ascii", 8, 12) === "WEBP" ? "webp" : undefined;
if (!extension) throw new Error("Unknown image MIME; inspect before saving");
await writeFile("image." + extension, bytes);
console.log(result.api429?.outputs);

The examples are checked offline against the client contract and public schema. No paid execution or result quality is claimed. A verified complete image jobs contract is not available here: asynchronous fields and polling are omitted.

n8n: HTTP Request and saving the result

1. Check access to the exact ID using GET /v1/models.
2. Create an HTTP Request with POST to https://gateway.api429.com/v1/images/generations; store the Bearer key in credentials.
3. Send the JSON from curl with one result. Do not include the key in a workflow export.
4. Parse JSON, check the data count and warnings; decode b64_json and check MIME before saving.

HTTP Request documentation. Compatibility with specialized AI nodes and native SDKs needs separate verification.

Three options for your own comparison

Compare identical tasks and acceptance criteria. These are image models, not a quality ranking; conditional rows apply to their own configurations and do not form an equivalent benchmark.

Version and release date

Manufacturer release date, OpenAI: Apr 21, 2026. The exact predecessor has not been established in the reviewed sources. This does not automatically replace the ID. The model’s addition date in API429 is unknown.

Frequently asked questions

Sources and review

• developers.openai.com/api/docs/models/gpt-image-2
• developers.openai.com/api/docs/guides/image-generation
• developers.openai.com/api/docs/pricing
• developers.openai.com/api/docs/changelog
• gateway.api429.com/api/public/catalog
• gateway.api429.com/api/public/openapi.json
• client.api429.com/documentation

Editorial review: Oct 24, 2026. Pricing · Getting started · Compare the family: GPT Image