gpt-image-2 API pricing: $8 input, $30 output
- OpenAI’s API pricing page now lists gpt-image-2 at $8 per 1 million image input tokens, $2 cached, and $30 per 1 million image output tokens after its April 21 launch. - OpenAI also says developers can call gpt-image-2 through both the Image API and Responses API, while fal.ai markets the same model for text-heavy graphics, packaging, and product photography. - The new model undercuts GPT Image 1’s $10 input and $40 output image-token rates while replacing chatgpt-image-latest as OpenAI’s recommended API image model. (openai.com)
OpenAI now prices gpt-image-2 at $8 per 1 million image input tokens and $30 per 1 million image output tokens. (openai.com) The same pricing page lists cached image input at $2 per 1 million tokens and text input at $5 per 1 million tokens. OpenAI added gpt-image-2 to its multimodal model lineup after the model’s April 21, 2026 release. (openai.com 1) (openai.com 2) OpenAI’s developer docs say gpt-image-2 can be used through the Image API for one-shot generations and edits, or through the Responses API for multi-turn image workflows. The docs also say some organizations may need API Organization Verification before using GPT Image models. (developers.openai.com) In plain terms, image tokens are the meter for pictures the model reads and pictures it produces, while the API endpoints are the pipes that move those requests. That pricing matters most for teams generating large batches of ads, mockups, or edited assets instead of a few single images. (openai.com) (developers.openai.com) The price change is clearer against OpenAI’s older image models. GPT Image 1 is listed at $10 per 1 million image input tokens and $40 per 1 million image output tokens, while chatgpt-image-latest shows $8 input and $32 output. (developers.openai.com 1) (developers.openai.com 2) That puts gpt-image-2 below GPT Image 1 on both sides of the image-token bill and slightly below chatgpt-image-latest on output pricing. OpenAI’s model page for chatgpt-image-latest now says it recommends gpt-image-2 for API use. (openai.com) (developers.openai.com) Fal.ai, which hosts the model on its own platform, says GPT Image 2 launched April 21 and pitches it for photorealism, text rendering, and brand-consistent product photography. Its model page highlights text-to-image and image-editing endpoints built around the same OpenAI model. (fal.ai) Fal.ai’s examples focus on packaging, signage, infographics, and electronic commerce catalog work, where legible labels and repeatable layouts are more valuable than novelty alone. Those are the kinds of workloads where per-token costs become a budget line, not a rounding error. (fal.ai) The short version is that OpenAI has attached a clearer production price to its newest image model and positioned it as the default API choice. For developers comparing model bills this week, the headline number is simple: $8 in, $30 out. (openai.com)
