gpt-image-1 API: what to call instead
OpenAI shut gpt-image-1 down on 2026-10-23. The gpt-image-1 API calls you have keep working here with one string changed: 5 GPT Image models on one key, priced per image and read live. Paid by card or in USDT, USDC, BTC, ETH, SOL, no KYC, nothing added on the top-up.
GPT Image models, per image
USD per image, read from the catalog at render. Up to three tiers per model; the media models page has every quality and resolution. OpenAI's list where the catalog has it.
| Model | Setting | Per image | OpenAI's list |
|---|---|---|---|
| GPT Image 2.5 Sunburstgpt-image-2.5-sunburst | quality low, resolution 1K | $0.0262 | — |
| quality low, resolution 2K | $0.053 | — | |
| quality low, resolution 4K | $0.0398 | — | |
| GPT Image 2.5 Flaregpt-image-2.5-flare | quality low, resolution 1K | $0.0262 | — |
| quality low, resolution 2K | $0.053 | — | |
| quality low, resolution 4K | $0.0398 | — | |
| GPT Image 2gpt-image-2 | quality low, resolution 1K | $0.024 | — |
| quality low, resolution 2K | $0.036 | — | |
| quality low, resolution 4K | $0.066 | — | |
| GPT Image 1.5gpt-image-1.5 | any settings | $0.0125 | — |
| GPT-4o Imagegpt-4o-image | any settings | $0.06 | $0.10 (40% under) |
The charge is fixed when the request is accepted and refunded in full if the generation fails. Each -edit twin bills the same tiers.
The same images call, one id changed
The endpoint is OpenAI's images API. Set the base URL, swap gpt-image-1 for an id in the table, keep the rest.
Python, OpenAI SDK
from openai import OpenAI client = OpenAI(base_url="https://api.relayswitch.io/v1", api_key="YOUR_RELAYSWITCH_KEY")result = client.images.generate( model="gpt-image-2", # was: gpt-image-1 prompt="A ceramic mug on a walnut desk, soft window light, product shot", size="1024x1024", quality="medium",)print(result.data[0].url)curl
curl https://api.relayswitch.io/v1/images/generations \ -H "Authorization: Bearer YOUR_RELAYSWITCH_KEY" \ -H "Content-Type: application/json" \ -d '{"model": "gpt-image-2", "prompt": "A ceramic mug on a walnut desk", "size": "1024x1024", "quality": "medium"}'size maps to the model's aspect ratio and resolution; quality picks the tier. The answer carries the file URL, served for 30 days.
Edits: the -edit twin on the generations endpoint
curl https://api.relayswitch.io/v1/generations \ -H "Authorization: Bearer YOUR_RELAYSWITCH_KEY" \ -H "Content-Type: application/json" \ -d '{"model": "gpt-image-2-edit", "input": {"prompt": "Replace the mug with a glass of water", "image_urls": ["https://example.com/mug.png"], "quality": "medium"}}'Picking a replacement for gpt-image-1
Least code change: GPT Image 1.5. One price per image, the same size strings (1024x1024, 1024x1536, 1536x1024), an -edit twin with mask_url. Change the model id and nothing else.
Same quality knob: GPT Image 2. quality low, medium or high, plus resolution 1K, 2K or 4K, each combination priced in the table. Your existing quality values still work; add resolution when you want more than 1K.
What OpenAI recommends: GPT Image 2.5 Sunburst, named on the gpt-image-1 model page as the current generation and editing model (read 2026-10-27), with Flare as the lighter build. Both add background and output_format and a quality range up to max.
OpenAI's image generation guide documents the request fields; the catalog at GET https://api.relayswitch.io/v1/catalog/media lists the values each id accepts here and the price of each.
What is stored: the email address, the top-ups, the keys, and one generation record per request with its file for 30 days. Not stored: a name, an ID, a phone number, a card for crypto top-ups. Details on the privacy page.
gpt-image-1 and its successors: questions
Availability, pricing and the swap.
GPT Image on a prepaid key
From $10 by card or in USDT, USDC, BTC, ETH, SOL. No OpenAI account, no KYC.