OpenAI sets image API at $0.008

Image APIs usually get expensive in two ways. The first is obvious — every output costs money. The second is sneakier — the bill gets harder to predict once size, quality, and editing passes all price differently. That is why this GPT Image 2 story matters. OpenAI’s latest image model is now being sold through Atlas Cloud at $0.008 per image, and the pitch is basically that the pricing stays flat enough to make experimentation feel cheap again. ### What actually launched? The model is GPT Image 2 — OpenAI’s current flagship image-generation model for fast, high-quality generation and editing. OpenAI lists it as supporting both image generation and image edits, and Atlas Cloud has wrapped those into two model IDs: one for text-to-image and one for editing an existing image. ### Where does the $0.008 number come from? (atlascloud.ai) The catch is that $0.008 is not OpenAI’s direct headline list price. Atlas says the model’s list price on its platform is $0.01 per image, then applies a 20% discount to get to $0.008. So the news here is really about distribution and packaging — OpenAI made the model available, and Atlas is marketing a cheaper effective rate on top. (developers.openai.com) ### Is OpenAI itself charging per image? Not exactly in the simple way the headline suggests. OpenAI’s official pricing page lists GPT-Image-2 in token terms — $8 per 1M image input tokens, $2 per 1M cached image input tokens, and $30 per 1M output tokens, plus text input token charges. OpenAI’s docs also say older GPT Image models still have a legacy per-image table, which implies GPT Image 2 is part of a newer pricing setup that is more token-based under the hood. (atlascloud.ai) ### So why does the flat reseller price matter? Because most developers do not want to estimate output tokens for every mockup, ad variant, or edit pass. They want to know, “If I generate 1,000 images, what will that cost?” Atlas is selling that answer directly. It says the same billed rate covers text-to-image and edits, and that the price stays flat across three quality tiers and multiple output sizes. That turns a fuzzy metered service into something closer to a utility. (openai.com) ### Why are edits the interesting part? Edits are where image workflows usually sprawl. You make a base image, fix the text, change the background, swap the product color, then crop for another format. Atlas argues that when edits cost the same as generations, those multi-step pipelines get much cheaper to test. Its example is blunt — several edit passes can still cost less than one call to some competing image models on its marketplace. (atlascloud.ai) ### How capable is GPT Image 2 supposed to be? OpenAI positions GPT Image 2 as its state-of-the-art image model, with support for flexible sizes, image inputs, and strong editing. Atlas adds the practical angle — strong prompt following and good text rendering inside images, which is a big deal for UI mockups, ads, and product visuals. But Atlas also says it is not the best choice for every style, especially heavily stylized 3D looks or ultra-photoreal portrait work. (atlascloud.ai) ### Does this change the market? At least a little. Atlas says GPT Image 2 is now its cheapest paid image API, below models like Flux 2 Pro, Imagen 4 Ultra, and Nano Banana 2. That does not prove OpenAI is cheapest everywhere, but it does show where the pressure is moving — away from premium per-image pricing and toward low-cost, high-volume creative tooling. (developers.openai.com) ### Bottom line? The real story is not that OpenAI suddenly made image generation universally cost $0.008. It is that GPT Image 2 is now being packaged by a reseller in a way that strips out pricing complexity. For builders, that may be the more important innovation anyway. (openai.com) (atlascloud.ai)