How much does GPT Image 2.6 cost per image?

It depends on the inputs and the output size. Our measured costs were $0.060 for a one-image edit, $0.071 for a 16:9 thumbnail and $0.088 to $0.108 for a 4:5 ad with five reference images.

Is GPT Image 2.6 Flare cheaper than Sunburst?

No. Both models have the same token prices. Flare returned slightly more output tokens in our tests, about $0.001 more per ad.

Does the complexity setting cost extra?

No. OpenAI states that adjusting complexity has no separate fee.

Is GPT Image 2.6 more expensive than GPT Image 2.5?

No. The token prices are identical, and 2.5 Sunburst and 2.6 Sunburst used exactly the same number of tokens for the same ad in our test.

GPT Image 2.6 pricing: what one image actually costs

Updated 5 November 2026

GPT Image 2.6 has no flat price per image. OpenAI charges $8 per million image input tokens, $5 per million text input tokens and $30 per million image output tokens, the same for Sunburst and Flare and the same as GPT Image 2.5. Unlike 2.5, there is no half-price Batch option for 2.6 yet.

In our tests that came to $0.060 for a one-image edit, $0.071 for a 16:9 thumbnail with four reference images and $0.088 to $0.108 for a 4:5 Meta ad with five reference images. About half of each ad's cost was the output image and the rest was the pictures we sent in.

GPT Image 2.6 prices

These are OpenAI's published rates for both 2.6 models. The complexity setting carries no separate fee.

Token type	Price per 1M tokens
Image input	$8.00
Text input	$5.00
Image output	$30.00

The model pages also list cached rates of $2 per million for images and $1.25 for text, but OpenAI's pricing page says cached image rates only apply through the Responses API, which the 2.6 models do not support. Plan on the full rates.

How the cost of one image is worked out

Every response reports the tokens it used, split into image input, text input and output. Multiply each by its rate and add them up. Here is our 4:5 Virlo ad on Sunburst 2.6, with five reference images and a detailed prompt:

Image input: 4,577 tokens x $8 per million = $0.0366
Text input: 790 tokens x $5 per million = $0.0040
Image output: 1,587 tokens x $30 per million = $0.0476
Total: $0.0882 for one 1088 x 1360 ad at high quality

The output part depends on the size and quality you ask for. The input part depends on how many images you send and how big they are: our five reference images came to 4,577 tokens, and a single 1088 x 1360 image plus a short prompt came to 1,526.

What our images cost

Measured on 5 November 2026, quality high, identical inputs for each model.

Job	Tokens in / out	Sunburst 2.6	Flare 2.6
4:5 ad, 5 reference images (dashboard)	5,367 / about 1,600	$0.088	$0.089
4:5 ad, 5 reference images (packaging)	7,695 / about 1,600	$0.107	$0.108
16:9 thumbnail, 4 reference images	5,042 / about 1,080	$0.071	$0.071
One-line edit, 1 image	1,526 / about 1,600	$0.060	$0.061

GPT Image 2.5 Sunburst used exactly the same tokens as 2.6 Sunburst on the two ads, so it cost the same to the cent.

What 100 ads cost

At our settings, 100 ads in 4:5 cost $8.80 to $10.80 in OpenAI fees, depending on how many and how large the reference images are. Making the 1:1 and 9:16 versions of each ad from the finished 4:5 adds a call per size.

A plain text-to-image request with no reference images would skip the input image cost. Based on the output tokens we measured, we estimate about $0.05 for a 1088 x 1360 image at high quality, though we have not measured that case directly.

What each quality level costs

OpenAI publishes no per-image table for 2.6, but its image guide includes a calculator. Using it, eesel estimates the output cost of one 1024 x 1024 image on 2.6 at about $0.025 on low, $0.044 on medium, $0.053 on high, $0.094 on xhigh and $0.211 on max. These are estimates, not measurements: our measured high-quality output at the larger 1088 x 1360 came to $0.048. The gap between high and max is the biggest lever on cost.

How 2.6 compares with older OpenAI models

Per-token prices have not changed since gpt-image-2, but the number of output tokens has. In September we measured the same kind of ad, at 1024 x 1536 and high quality, on gpt-image-2 and GPT Image 2.5.

Model	Output tokens	Cost per ad	Measured
gpt-image-2	5,488	$0.197	September 2026, 1024 x 1536
GPT Image 2.5 Sunburst	1,372	$0.073	September 2026, 1024 x 1536
GPT Image 2.6 Sunburst	1,587	$0.088	November 2026, 1088 x 1360
GPT Image 2.6 Flare	1,629	$0.089	November 2026, 1088 x 1360

The November ads used a larger output size and different reference images, so the rows are not a strict like for like. The step that cut the price was 2.5, which used about a quarter of gpt-image-2's output tokens. 2.6 kept that saving.

How to spend less per image

Send fewer reference images, and only the ones the model needs to see.
Shrink large photos before you send them. We cap the long side at 1536 pixels.
Ask for the size you will use, not a bigger one, then scale down.
Use complexity low for small edits: it cost the same and finished about 15 seconds faster on Sunburst in our test.
Pick Sunburst or Flare on speed and quality, not price. Their prices are the same.
For large jobs that can wait up to a day, GPT Image 2.5 through the Batch API costs half its standard rate. 2.6 has no Batch price yet.

Frequently asked questions

How much does GPT Image 2.6 cost per image?: It depends on the inputs and the output size. Our measured costs were $0.060 for a one-image edit, $0.071 for a 16:9 thumbnail and $0.088 to $0.108 for a 4:5 ad with five reference images.
Is GPT Image 2.6 Flare cheaper than Sunburst?: No. Both models have the same token prices. Flare returned slightly more output tokens in our tests, about $0.001 more per ad.
Does the complexity setting cost extra?: No. OpenAI states that adjusting complexity has no separate fee.
Is GPT Image 2.6 more expensive than GPT Image 2.5?: No. The token prices are identical, and 2.5 Sunburst and 2.6 Sunburst used exactly the same number of tokens for the same ad in our test.