Nano Banana 2.1 and GPT Image 2.5 both list $30 per million tokens

Google's Nano Banana 2.1 standard image output is $30.00 per million tokens. OpenAI's GPT Image 2.5 standard image output is also $30 per million tokens. The per-token rate is the same, so the cost per image depends on how many tokens each model spends per image. Google publishes per-image equivalents: $0.0336 (1K), $0.0504 (2K), $0.113 (4K). Dividing by $30 per million gives 1,120, 1,680 and about 3,767 tokens.

What OpenAI's page does and does not say

The OpenAI pricing page lists token rates for gpt-image-2.5: image output $30.00 per million, image input $8.00, text input $5.00 (batch image output $15.00). It does not list per-image prices; it points to a calculator in the image generation guide. So no per-image OpenAI figure is claimed from the vendor page here.

Sume's Image API docs quote the same $30 per million output image tokens for gpt-image-2.5 via its provider and say that at 1024x1024, xhigh output is $0.09366 before input tokens and Sume pricing, which implies about 3,122 tokens ($0.09366 / $30 x 1,000,000).

What Sume charges per image

The Sume rows are per image and include Sume's margin.

How to read the Sume price

For Nano Banana 2.1, Sume's 1K price of $0.10 is 2.98 times Google's $0.0336 (0.10 / 0.0336), and the 4K price of $0.20 is 1.77 times Google's $0.113. Sume's rows are not a flat 1.25 multiple of Google's published per-image figures. I have not established the reason from the sources read here, so treat the Google page as the vendor's price and the Sume catalog as what Sume charges.

If you only need to know the price of a Sume image before you submit it, the catalog or GET /v1/images/models is the source of truth. The table is not a recommendation to use either model.