Image Model

GPT Image 2

OpenAI image generation pricing last checked on 2026-04-22. 1024x1024 images start at $0.006 and go up to $0.211 depending on quality.

Quick Answer

GPT Image 2is OpenAI's image generation and editing model. Based on the official OpenAI pricing and docs checked on 2026-04-22, text tokens cost $5.00 input, $1.25 cached input, and $10.00 output per 1M tokens.

Image tokens cost $8.00 input, $2.00 cached input, and $30.00 output per 1M tokens. A 1024x1024 image is listed at $0.006 on low quality, $0.053 on medium, and $0.211 on high.

Pricing Snapshot

Text tokens

$5.00 input · $10.00 output

Image tokens

$8.00 input · $30.00 output

1024x1024 image

Low $0.006 · Medium $0.053 · High $0.211

1024x1536 image

Low $0.005 · Medium $0.041 · High $0.165

1536x1024 image

Low $0.005 · Medium $0.041 · High $0.165

Transparent bg

Not supported

Model Details

Provider

OpenAI

Quality

low, medium, high, auto

Popular sizes

1024x1024, 1536x1024, 1024x1536

Snapshot

gpt-image-2-2026-04-21

Generation endpoint

/v1/images/generations

Edit endpoint

/v1/images/edits

Max edge

3840px

Pixel range

655,360-8,294,400

GPT Image 2 supports sizes up to 3840px, requires a max aspect ratio of 3:1, and adds 100 extra image output tokens per streamed partial image.

FAQ

How much does GPT Image 2 cost per image?

According to OpenAI's pricing checked on 2026-04-22, a 1024x1024 image costs $0.006 on low quality, $0.053 on medium, and $0.211 on high.

What sizes and quality levels does GPT Image 2 support?

GPT Image 2 supports low, medium, high, auto quality modes and popular sizes such as 1024x1024, 1536x1024, 1024x1536. The maximum edge length is 3840px and the aspect ratio limit is 3:1.

Does GPT Image 2 support image edits?

Yes. OpenAI documents GPT Image 2 on both /v1/images/generations for image generation and /v1/images/edits for image edits.

Does GPT Image 2 support transparent backgrounds?

No. OpenAI currently documents transparent backgrounds as unsupported for GPT Image 2.

Recent Changes

No editorial changelog entries for this model yet.