GPT Image 2 API Price per Image: 23 Providers Compared
8 min read · Published · Updated
APIMODELS, our own product, publishes a 30-day success rate on every model page, 99.9% for gpt-image-2; of the other 22 GPT Image 2 API providers compared, only OpenRouter and CometAPI publish any model-level reliability figure. Prices were read on November 11, 2026, and the lowest medium 1K price is APIMODELS at 0.025 USD.
Disclosure
APIMODELS is our own product: it is made by the team behind GenGrades, and GenGrades' content backend runs on hub.apimodels.app. It ranks first because it has the lowest stated price for a medium-quality 1K image, the sort key used for providers that state quality. Its column is built from its public pages like the others, its success rates are its own figures, and the table shows where other providers charge less.
Side by side
Every pick in its own column, compared on the same rows: price, fees, free use, models, MCP, Skills and more.
What each one is, who it suits, and the full review.
1. APIMODELS
Medium-quality 1K and 2K images, GPT Image 2.6, paying by Alipay or USDTOur product
APIMODELS, made by the team behind GenGrades, sells gpt-image-2 at medium quality for 0.025, 0.03 and 0.05 USD at 1K, 2K and 4K, with low and high on its gpt-image-2-all route. It needs no OpenAI account, takes cards, PayPal, Alipay and USDT, and publishes a 30-day success rate on each model page.
LaoZhang API charges 0.03 USD per image on its gpt-image-2-vip route at every size from 1K to native 4K and at low, medium or high quality, and the same for GPT Image 2.6 on its -vip routes. A token-billed group passes OpenAI's rates through.
WeChat Pay, Alipay or USDT, with a choice of billing routes
APIYI bills gpt-image-2 by tokens at OpenAI's rates and also sells two per-call routes that its own docs describe as reverse-engineered rather than OpenAI's API: gpt-image-2-all at 0.025 USD (about 1K, no quality setting) and gpt-image-2-vip at 0.03 USD (1K to 4K). It takes WeChat Pay, Alipay, USDT and cards.
PiAPI bills gpt-image-2 by tokens at 30 USD per 1M output tokens, accepts only medium quality and estimates that at about 0.032 USD for 1024×1024. Its landing page shows 0.10 USD per image, the price its docs give for the separate gpt-image-2-preview route. New accounts get 0.50 USD of credit; GPT Image 2.6 is not offered.
PoYo prices gpt-image-2 per image in credits: 0.010, 0.0424 and 0.1688 USD for low, medium and high at 1K, and 0.0808 USD for medium at 4K. It also sells GPT Image 2.6 on a fixed route at 0.0085 USD at 1K with no quality choice.
Teams already on Replicate who want one price per quality at any size
Replicate charges gpt-image-2 per output image by quality alone, at any size: 0.012 USD low, 0.047 medium and 0.128 high. You can pass your own OpenAI key to pay OpenAI directly. Its GPT Image 2.6 prices run from 0.05 to 0.50 USD per image.
Runway's developer API sells gpt-image-2 in one-cent credits: 0.01, 0.05 and 0.20 USD for low, medium and high at 1K or 2K, and 0.02, 0.11 and 0.41 USD at 4K. Its default quality is high. GPT Image 2.6 medium is 0.09 USD at its 1K size.
EachLabs passes OpenAI's token cost through unchanged and lists it per image: 0.0059, 0.0527 and 0.2108 USD for low, medium and high at 1024×1024, and 0.0048 USD for low at 2048×1152. GPT Image 2.6 is listed at 0.0436 USD for medium at 1024×1024.
OpenAI bills gpt-image-2 and GPT Image 2.6 by tokens at the same rates; its calculator puts a 1024×1024 gpt-image-2 image at 0.006 USD low, 0.053 medium and 0.211 high. Its image guide says you may need API Organization Verification first, with an original government ID.
fal bills gpt-image-2 per token at OpenAI's rates and lists the per-image results, 0.006, 0.053 and 0.211 USD at 1024×1024. Its API defaults to high quality. You can pass your own OpenAI key.
WaveSpeedAI prices gpt-image-2 per image: 0.01, 0.06 and 0.22 USD for low, medium and high at 1K, with 5% off shown at the Generate button. Eligible new accounts get 1 USD of credit.
Segmind bills gpt-image-2 by tokens and publishes sample per-image prices: 0.01, 0.06 and 0.22 USD for low, medium and high at 1024×1024, and 0.11 USD for medium at 3840×2160. Its rate table lists 31.58 USD per 1M output tokens.
Apiframe states 0.06 USD for gpt-image-2 at medium 1024×1024 and only a 0.02–0.16 USD range for other settings. Output goes up to 2K, not 4K, top-ups need a paid plan from 39 USD a month, and GPT Image 2.6 is not listed.
The lowest flat 1K price, when the quality tier does not matter
APIMart lists gpt-image-2 at 0.0085 USD per image at 1K, 0.015 at 2K and 0.022 at 4K, with no quality setting and no statement of which channel the route uses; a token-billed gpt-image-2-official model is also offered. GPT Image 2.6 is 0.0085 USD at 1K on its -ext route. It takes cards, Alipay, crypto and PayPal.
Kie.ai sells gpt-image-2 as an async task API priced by resolution only: 0.03, 0.05 and 0.08 USD at 1K, 2K and 4K, with no quality setting listed. Failed tasks are not charged and new users get 80 free credits.
Together AI lists gpt-image-2 at a flat 0.053 USD per image with 1K, 2K and 4K output and no quality setting stated. It offers no free trial and requires a 15 USD first purchase; GPT Image 2.6 is not on its price list.
CometAPI bills gpt-image-2 by tokens at 4 USD per 1M input and 24 USD per 1M output, 20% below OpenAI, and GPT Image 2.6 at the same discount. New accounts get 0.10 USD of credit. A page widget shows 97.3%, labelled both as 24-hour uptime and as a 30-day success rate.
Token billing 10% below OpenAI, paid by Alipay or WeChat Pay
EvoLink bills gpt-image-2 by tokens at 27 USD per 1M output tokens, 10% below OpenAI, and sells a gpt-image-2-beta route at 0.015 USD per 1K image that it says is not an official OpenAI model ID. GPT Image 2.6 is at OpenAI's rates. It takes cards, Alipay, WeChat Pay and bank transfer.
Atlas Cloud bills gpt-image-2 by tokens at OpenAI's rates but shows a per-run estimate of about 0.004 USD without saying which quality and size it assumes; GPT Image 2.6 shows about 0.007 USD. It takes cards, Alipay, WeChat Pay and crypto, with a 25 USD minimum top-up.
Teams already on OpenRouter who want live availability figures
OpenRouter passes OpenAI's token rates through for gpt-image-2 and both GPT Image 2.6 tiers, billed in OpenRouter credit. It publishes live availability and latency per model; each of these models has a single upstream, OpenAI.
Vercel AI Gateway serves gpt-image-2 and both GPT Image 2.6 tiers at OpenAI's token rates, billed through Vercel or through your own key with no added fee. Its 5 USD monthly free credit does not cover these models.
Azure offers gpt-image-2 as generally available with no application form, at OpenAI's token rates on a Global deployment, in five standard regions with a default quota of 5 images per minute. GPT Image 2.6 is not listed yet.
Cloudflare serves gpt-image-2 at OpenAI's token rates through your own OpenAI key or Unified Billing credits, which carry a 5% fee on purchase. Transparent backgrounds are not supported and GPT Image 2.6 is not in its catalog.
Stated plainly, with where each one comes from and when we last looked.
Claim
Value
Source
OpenAI will remove gpt-image-2 and the gpt-image-2-2026-04-21 snapshot from the API on February 10, 2027, and names gpt-image-2.6-sunburst or gpt-image-2.6-flare as replacements.
APIYI's documentation describes its gpt-image-2-all (0.025 USD per call) and gpt-image-2-vip (0.03 USD per call) routes as reverse-engineered routes rather than OpenAI's API.
This page lists what one gpt-image-2 image costs at 23 ways to buy it through an API: OpenAI and Azure directly, and 21 services that resell or route it. Every price was read on the provider's own page on November 11, 2026, and each cell links to it. APIMODELS is our own product, as the note at the top says. GenGrades has not run these providers; this is a price comparison, not a test.
Reliability: who publishes numbers
Price is only half of the choice: an endpoint that fails often costs more in retries and lost work than it saves per image. Few providers publish reliability figures for gpt-image-2, and they measure different things:
Provider
Published figure
What it measures
APIMODELS
99.9% for gpt-image-2
30-day success rate per model, October 12 to November 11, 2026, excluding content-policy refusals and malformed requests
OpenRouter
99.68%
3-day availability for gpt-image-2 on November 11, 2026
CometAPI
97.3%
Labelled both as 24-hour uptime and as a 30-day success rate
Platform uptime claims or SLAs, with no model-level data
The other 14 providers
None published
All of these figures are self-reported. APIMODELS shows the rate on each model page, does not bill failed calls, and switches to another channel when one fails.
How the list is ordered
Prices are not comparable unless quality and size match, so the table has three groups. First come providers that state the quality tier, ordered by the price of a medium-quality 1K image: APIMODELS at 0.025 USD, LaoZhang and APIYI at 0.03, PiAPI at about 0.032, then PoYo, Replicate, Runway, EachLabs, OpenAI, fal, WaveSpeedAI, Segmind and Apiframe. Next come flat prices that state no quality tier: APIMart, Kie.ai and Together AI. Last come providers that bill only by tokens: CometAPI, EvoLink, Atlas Cloud, OpenRouter, Vercel AI Gateway, Azure and Cloudflare.
The lowest price depends on quality and size
Image
Lowest price found
Where
1K, low
0.0059 USD (1024×1024)
EachLabs; OpenAI and fal ≈0.006
1K, medium
0.025 USD
APIMODELS
1K, high
0.03 USD
LaoZhang and APIYI per-call routes
1K, quality not stated
0.0085 USD
APIMart
2K, medium
0.03 USD
APIMODELS, LaoZhang, APIYI
4K, medium or high
0.03 USD
LaoZhang and APIYI per-call routes
4K, quality not stated
0.022 USD
APIMart
Token billing favours wide sizes: OpenAI's calculator prices 2048×1152 at medium at about 0.042 USD, against 0.107 USD for 2048×2048. Token-billed figures here count output tokens only, so long prompts and reference images add to them.
Flat per-call routes: check what you are buying
Several of the lowest prices come from per-call routes that charge one price whatever the size or quality. Read the provider's own description before relying on one. APIYI's docs describe its gpt-image-2-all and gpt-image-2-vip routes as reverse-engineered rather than OpenAI's API. EvoLink says its 0.015 USD gpt-image-2-beta is not an official OpenAI model ID. APIMart and Kie.ai offer no quality setting and do not say which tier their output matches. These routes can suit drafts and volume work, but output size, quality and availability may differ from OpenAI's API.
Do you need OpenAI organization verification?
Only if you call OpenAI directly or bring your own OpenAI key. OpenAI's image guide says you may need to complete API Organization Verification before using GPT Image models; its help center asks for an original, physical government-issued ID, allows one organization per person and says results cannot be manually overridden. Every reseller here bills in its own credit, and none of their pages mention verification. Azure needs an Azure subscription instead.
Paying with Alipay, WeChat Pay or USDT
Among the providers whose payment pages we read, APIMODELS takes Alipay and USDT, APIYI takes WeChat Pay, Alipay and USDT, Atlas Cloud takes Alipay, WeChat Pay and crypto, APIMart takes Alipay, crypto and PayPal, and EvoLink takes Alipay and WeChat Pay. The large Western platforms bill cards in US dollars.
gpt-image-2 shuts down on February 10, 2027
OpenAI deprecated gpt-image-2 and its snapshot gpt-image-2-2026-04-21 on November 10, 2026, removes them from the API on February 10, 2027, and names gpt-image-2.6-sunburst and gpt-image-2.6-flare as replacements. Every provider here depends on OpenAI's model, so plan the move to 2.6 whichever one you use. Of the ten provider pages we checked for it, only APIMODELS mentions the date.
GPT Image 2.6 prices
OpenAI charges GPT Image 2.6 the same token rates as gpt-image-2, but at 1024×1024 its calculator gives 0.025 USD low, 0.043 medium and 0.053 high, against 0.006, 0.053 and 0.211 for gpt-image-2. Among resellers, APIMart and PoYo sell fixed 2.6 routes at 0.0085 USD at 1K with no quality choice, APIMODELS charges 0.01 USD low and 0.02 medium at 1K, LaoZhang 0.03 at any quality, and WaveSpeedAI 0.05 for medium. Azure, Cloudflare, Together AI, PiAPI and Apiframe do not list 2.6.
Who should go direct to OpenAI or Azure
Go direct if you can pass organization verification and need OpenAI's own terms, or the Batch API, which halves gpt-image-2 rates. Choose Azure if your company buys through a Microsoft agreement. Any reseller adds a company between you and OpenAI: read its terms, refund policy and data handling before sending production traffic.
Where these tools fall short
What they still get wrong, with the evidence, and what no public test has settled yet.
Weak
Weak
GenGrades has not run these providers.
Evidence: Every price comes from the providers' own pages on November 11, 2026; no output quality, speed or reliability was measured by GenGrades.
Flat-price routes may not match OpenAI's output.
Evidence: APIYI and EvoLink describe their cheapest routes as unofficial, and APIMart and Kie.ai offer no quality setting, so their images may differ in size or quality from OpenAI's API.
Token-billed prices are estimates.
Evidence: Per-image figures for token-billed providers use OpenAI's calculator, which counts output tokens only; prompts and reference images cost extra.
Uncertain
Uncertain
Which quality APIMart, Kie.ai and Together AI deliver.
Why unsure: Their gpt-image-2 pages state one price and no quality tier.
What Atlas Cloud's ≈0.004 USD estimate covers.
Why unsure: The page bills by OpenAI's token rates but does not say which quality and size the estimate assumes; its quote tool needs a login.
What happens to these endpoints after February 10, 2027.
Why unsure: Most provider pages do not mention OpenAI's shutdown or say whether they will remove or redirect the model.
Common questions
Answer first, then why, an example, the limits and what to do next.
Which GPT Image 2 API is the cheapest?
The cheapest provider depends on quality and size. At 1K, EachLabs is lowest for low quality at 0.0059 USD (1024×1024), APIMODELS, our own product, for medium at 0.025 USD, and LaoZhang and APIYI for high at 0.03 USD on per-call routes. APIMart lists 0.0085 USD at 1K with no quality setting. At 4K, LaoZhang and APIYI charge 0.03 USD at any quality. Prices were read on November 11, 2026.
How much does GPT Image 2 cost per image?
OpenAI's own price for a 1024×1024 gpt-image-2 image is about 0.006 USD at low, 0.053 at medium and 0.211 at high quality, billed by tokens. Resellers range from 0.0085 USD for a flat 1K image to 0.06 USD for medium at 1K, and flat per-call routes charge 0.03 USD up to 4K.
Do I need OpenAI organization verification to use GPT Image 2?
Verification applies only when you call OpenAI's API directly or bring your own OpenAI key. OpenAI's image guide says you may need API Organization Verification before using GPT Image models, with an original government ID and one organization per person. Every reseller in this comparison bills in its own credit and does not mention verification.
Which GPT Image 2 API can I pay for with Alipay or USDT?
APIMODELS takes Alipay and USDT, APIYI takes WeChat Pay, Alipay and USDT, Atlas Cloud takes Alipay, WeChat Pay and crypto, APIMart takes Alipay, crypto and PayPal, and EvoLink takes Alipay and WeChat Pay, according to their own pages on November 11, 2026.
Is GPT Image 2 being discontinued?
OpenAI will remove gpt-image-2 and the gpt-image-2-2026-04-21 snapshot from its API on February 10, 2027, as announced on November 10, 2026, and names GPT Image 2.6 Flare and Sunburst as replacements. Every reseller depends on OpenAI's model, so their gpt-image-2 endpoints are affected too.
Which GPT Image 2 API provider is the most reliable?
A fair reliability ranking is not possible, because only a few providers publish numbers and they measure different things. APIMODELS shows a 30-day success rate of 99.9% for gpt-image-2, OpenRouter showed 99.68% three-day availability on November 11, 2026, and CometAPI shows 97.3% with an unclear label. All are self-reported; most providers publish only platform uptime claims or nothing.
First version: 23 GPT Image 2 API providers compared on price per image by quality and size, OpenAI verification, GPT Image 2.6 and published reliability data, each cell read on the provider's own page.
GPT Image 2.6 is OpenAI's image generation and editing model, released November 4, 2026 as gpt-image-2.6-flare for fast everyday images and gpt-image-2.6-sunburst for detailed work and precise edits. It adds animated GIF and SVG output, markup-guided edits and a complexity setting, and replaces gpt-image-2, which shuts down on February 10, 2027.
Claude Fable 5.2 is Anthropic's most capable public model, released November 9, 2026 at 8 and 40 USD per million input and output tokens, 20% below Fable 5.1. Against Opus 5.5 it costs double per token and, at max effort, scores 65 to 59 on Artificial Analysis's index for 26.54 USD per task against 10.04; Anthropic still recommends Opus first.
Hailuo AI is MiniMax's AI video and image app for web and Android, built since July 31, 2026 around MiniMax H3, an open-weight model that makes up to 15 seconds of 2K video with sound. Plans start at US$10.50 a month for 1,000 credits. Payments are not refunded, and a studio copyright suit over Hailuo survived dismissal in May 2026.
Devin is Cognition's autonomous coding agent: it works on its own cloud computer and opens pull requests, and ships with Devin Desktop (formerly Windsurf) and a CLI. In November 2026 Cognition added the cheaper SWE-2 Flash model and Devin Review 2.0. Plans cost US$0 to US$200 a month; the best-known outside test, Answer.AI's, saw 3 successes in 20 tasks.