GPT Image 2.5 Selection & API Guide

Choosing a provider

Published quotes are compared by resolution without separate quality tiers. Check 1K results first, then choose a provider for your output size.

Confirm model aliases in each provider’s documentation; a similar alias is not proof of upstream identity. Samples are references, and results depend on prompts, settings and providers.

Verified 1K prices

See the price page for purchase conditions and complete specifications

Connect to ImgAPI

Set IMGAPI_CARD_KEY on the server. The example generates one 1K image and returns image URLs; ImgAPI bills the request.

curl --fail-with-body --max-time 300 \
  'https://imgapi.vip/prod-api/v1/images/generations' \
  -H "Authorization: Bearer $IMGAPI_CARD_KEY" \
  -H 'Content-Type: application/json' \
  --data '{
  "model": "gpt-image-2.5",
  "prompt": "A flower shop after rain.",
  "size": "1024x1024",
  "quality": "auto",
  "n": 1,
  "response_format": "url"
}'

ImgAPI documentation

Provider addresses and protocols

Swipe to view all columns

For compatible synchronous APIs, replace the base URL, key and provider model alias, then verify size, quality and response format. Asynchronous APIs require submitting a job and querying its result.

Troubleshooting

401: check the key. 403: check permissions. 404: check endpoint and model alias. 429: check credits and rate limits. Check provider records after a timeout or unclear response before submitting again.