Skip to content
Greatest Calculators

AI Image Generation Cost Calculator

See what one AI-generated image costs through each provider's API at the size and quality you need, and what your monthly volume costs. Token-priced and per-image models are converted to one price per image, from prices read on each provider's own page.

Cost per image

$0.0137per image

OpenAI GPT Image 2.5 (Sunburst or Flare) · 1024 × 1024 · medium · 1,000 images a month

Per month
$13.67
Per 1,000 images
$13.67
Per year
$164.04
Model, size and quality

Both GPT Image 2.5 models bill $30 per 1M image output tokens and $5 per 1M prompt tokens (Batch: half). Tokens per image are OpenAI's own estimate for each quality and size; OpenAI says actual usage can differ by model and image. Sizes above 2560 × 1440 are experimental. Read November 2, 2026.

1,000 a month is about 33 a day.

Prompt length and images you keep
tokens

About 75 words is 100 tokens. Billed by OpenAI and Google only.

%

Every attempt is billed. 50% means two tries per image you use.

How the price is built

Image output tokens
439
Image output at $30 per 1M tokens
$0.01317
Prompt: 100 tokens at $5 per 1M
$0.0005
Cost per generated image
$0.01367

Every preset at 1K (about 1 megapixel), cheapest first

Your prompt length, and 1,000 images a month. Providers size their tiers differently, so each row names the size it prices.

Model, size and qualityPer imagePer month
GPT Image 2.5 (Sunburst or Flare)OpenAI · 1024 × 1024 · low$0.00638$6.38
FLUX 3 Image KleinBlack Forest Labs · 1K$0.009$9.00
GPT Image 2.5 (Sunburst or Flare)OpenAI · 1024 × 1024 · medium · your choice$0.0137$13.67
Grok Imagine ImagexAI · 1K (1024 × 1024)$0.02$20.00
Gemini 3.1 Flash Lite Image (Nano Banana 2 Lite)Google · 1K (1024 × 1024)$0.0336$33.62
Nano Banana 2.1Google · 1K (1024 × 1024)$0.0337$33.75
Grok Imagine Image 2.0xAI · 1K · low$0.04$40.00
FLUX 3 ImageBlack Forest Labs · 1K$0.048$48.00
GPT Image 2.5 (Sunburst or Flare)OpenAI · 1024 × 1024 · high$0.0532$53.18
Grok Imagine Image 2.0xAI · 1K · medium$0.06$60.00
Gemini 3.1 Flash Image (Nano Banana 2)Google · 1K (1024 × 1024)$0.0672$67.25
GPT Image 2.5 (Sunburst or Flare)OpenAI · 1024 × 1024 · xhigh$0.0942$94.16
Gemini 3 Pro Image (Nano Banana Pro)Google · 1K (1024 × 1024)$0.135$134.60
GPT Image 2.5 (Sunburst or Flare)OpenAI · 1024 × 1024 · max$0.211$211.22

Price sources

Every price read on the provider's own page on November 2, 2026. Standard paid-tier API prices in US dollars; confirm with the provider before you commit.

What this calculator assumes
  • One generation returns one image, made from a text prompt. Reference images for editing are billed on top and not modelled.
  • OpenAI and Google bill image output tokens plus prompt tokens; xAI and Black Forest Labs bill per image and do not bill the prompt separately.
  • Google's Gemini 3 image models always think, and Google bills the thinking as text output: $7.50 per million tokens on Nano Banana 2.1, $3 on Gemini 3.1 Flash Image, $1.50 on Gemini 3.1 Flash Lite Image and $12 on Gemini 3 Pro Image (half in Batch where Batch exists). How many thinking tokens an image uses depends on the prompt, so the estimate leaves them out and a Google image costs more than shown; the usage figures on a test call give the real count.
  • OpenAI's tokens per image are its own estimate for each quality and size; real usage can differ by model and image. Quality set to auto is billed at the quality served.
  • Standard paid-tier direct-API prices in USD. Batch prices only where the provider publishes them. Free tiers, subscription apps, search grounding, streaming previews and regional uplifts are excluded.
  • Images you keep ÷ share kept = images generated; every generation is billed, kept or not.

How this is calculated

I teach data analysis and SQL on The Data Science Tutor.

Formulas, assumptions, and rounding are documented in our calculator methodology.

Built by Will Henschell, data scientist. Formula checked against OpenAI, Google, xAI and Black Forest Labs image pricing pages and OpenAI's image token estimator, verified . Editorial policy.

Disclaimer: A planning estimate from published prices and the inputs you enter, not a quote or a bill. Prices were read on each provider's own page on November 2, 2026 and change often; OpenAI's tokens per image are its own estimate, and real usage varies by image. Google bills the thinking tokens of its Gemini 3 image models as text output and the estimate leaves them out, so a Google image costs more than shown. Reference images, grounding, regional uplifts and taxes are not included. Confirm current prices with the provider before you commit.

Prices and Sources — Read November 2, 2026

OpenAI: $30 per million image output tokens and $5 per million prompt tokens for both GPT Image 2.5 models (Sunburst and Flare), half that in Batch, from OpenAI's pricing page; tokens per image from the estimator in OpenAI's image generation guide. Google: Nano Banana 2.1, Gemini 3.1 Flash Image, Gemini 3.1 Flash Lite Image and Gemini 3 Pro Image prices and tokens per resolution from the Gemini API pricing page. xAI: Grok Imagine Image 2.0 and Grok Imagine Image from their model pages. Black Forest Labs: FLUX 3 Image and FLUX 3 Image Klein from its pricing page. Google's thinking (text output) rates from the same page, read November 3, 2026. OpenAI's, Google's standard and xAI's image prices, and Google's tokens per image, are checked against those pages every morning; Black Forest Labs' prices, OpenAI's tokens per image, the prompt prices and Google's thinking and batch rates are re-read by hand each month.

Choosing a Model on Cost

Quality is the biggest lever inside OpenAI's pricing and size the biggest at Black Forest Labs, where a 4K FLUX 3 Image costs about 13 times its 1K price. Google's Gemini 3 Pro Image charges the same for 1K and 2K, so the larger size costs nothing extra on that model. The cheapest row is not always the right one: a model whose images you keep more often can cost less per usable image, which is what the keep-rate input measures. Run a small test batch, count the images you would actually use, and enter that share.

How This Is Calculated

How the Cost Is Calculated

For a token-priced model (OpenAI, Google), one image costs its output tokens × the image output price per million tokens, plus the prompt tokens × the text input price per million. For a per-image model (xAI, Black Forest Labs), it costs the provider's price for that size and quality. Batch prices replace the standard ones only where the provider publishes them. If you keep a share k of what you generate, you pay for images needed ÷ k generations, so the cost per image you keep is the generation cost ÷ k. Monthly cost is the generation cost × generations per month, and the annual cost is twelve months of it.

Formulas

Token-priced models: per image = image tokens × p_out ÷ 1,000,000 + prompt tokens × p_text ÷ 1,000,000

Per-image models: per image = the provider's price for that size and quality

Generations = images you keep ÷ keep rate; Monthly = per image × generations

  • image tokens output tokens for the size and quality, from the provider (tokens)
  • p_out, p_text image output and text input prices (dollars per million tokens)
  • keep rate share of generations you keep (decimal, above 0 up to 1)

Assumptions and defaults

  • GPT Image 2.5 at 1024 × 1024, medium quality, produces 439 image tokens, billed at $30 per million; prompt text at $5 per million. Source: OpenAI API pricing and image generation guide (token estimator), as of November 2, 2026.
  • Other presets use each provider's standard per-image or per-token price for the size and quality; batch prices only where the provider publishes them. Source: OpenAI, Google, xAI and Black Forest Labs pricing pages, as of November 2, 2026.

Inputs it refuses or treats specially

  • A keep rate of 0 or above 1, a negative price or prompt-token count, or a non-whole image count is refused.
  • Google's Gemini 3 image models also bill thinking tokens, which depend on the prompt and are not in the estimate.

Worked example

GPT Image 2.5, 1024 × 1024, medium quality with a 100-token prompt: 439 × $30 + 100 × $5 per million = $0.01367 an image, $13.67 for 1,000 images.

Checked by the test “default: GPT Image 2.5, 1024 × 1024, medium, 100 prompt tokens = $0.01367 an image, $13.67 for 1,000” in src/__tests__/rollout-batch-0013.test.ts.

Check it yourself

The calculation is the function priceGeneration(model, point, { batch, promptTokens }): ImageGenerationPrice | null in src/lib/calculators/aiImageCost.ts. These lines reproduce the example and print 13.67.

Show the JavaScript and Python

JavaScript

const perImage = (439 * 30 + 100 * 5) / 1e6;
console.log((perImage * 1000).toFixed(2));

Python

per_image = (439 * 30 + 100 * 5) / 1e6
print(f"{per_image * 1000:.2f}")

I teach data analysis and SQL on The Data Science Tutor.

Sources

Last verified November 2, 2026.

50 automated tests cover this calculator. How calculations are documented and checked.

Frequently Asked Questions

How much does it cost to generate one AI image?

From under a cent to over 60 cents, depending on the model, size and quality. At about 1 megapixel (1K) with a 100-token prompt: GPT Image 2.5 at low quality about $0.0064, FLUX 3 Image Klein $0.009, Grok Imagine Image $0.02, Nano Banana 2.1 about $0.034, Gemini 3 Pro Image about $0.135 and GPT Image 2.5 at max quality about $0.21. At 4K, FLUX 3 Image costs $0.607 an image.

How are image tokens turned into a price per image?

OpenAI and Google bill the tokens a model uses to draw the image, at a price per million tokens, plus the prompt's text tokens. A GPT Image 2.5 image at 1024 × 1024 and medium quality uses 439 output tokens: 439 × $30 ÷ 1,000,000 = $0.01317. A 100-token prompt at $5 per million adds $0.0005, for $0.01367. xAI and Black Forest Labs simply charge a price per image.

Why does quality change the price so much on GPT Image 2.5?

Higher quality settings make the model spend more output tokens on the same image. OpenAI's own estimate for a 1024 × 1024 image runs from 196 tokens at low to 7,024 at max, about 36 times as many, so the price rises from $0.00588 to $0.21072 before the prompt. Its estimate also gives a 1536 × 1024 image fewer tokens than a square one at every quality; check your own usage figures if the difference matters.

Is batch image generation cheaper?

Where it is offered, yes: OpenAI's GPT Image 2.5 and Google's Nano Banana 2.1 and Nano Banana 2 Lite bill batch jobs at half the standard token prices, in exchange for asynchronous results. Gemini 3.1 Flash Image, Gemini 3 Pro Image and Black Forest Labs list no batch price, and xAI bills batch image jobs at its standard price, so the calculator keeps the standard price and says so.

Is a 4K image the same size at every provider?

No. Google's 4K is 4096 × 4096, about 16.8 megapixels. OpenAI's largest size is 3840 × 2160, about 8.3 megapixels, and it calls anything above 2560 × 1440 experimental. xAI's Grok Imagine models stop at 2K, and Black Forest Labs names its tiers by approximate megapixels (1K about 1 MP, 2K about 4 MP, 4K about 16 MP). The comparison table shows the size behind each row.

What is not included in the estimate?

Google's thinking tokens: Google's Gemini 3 image models always think, and Google bills the thinking as text output: $7.50 per million tokens on Nano Banana 2.1, $3 on Gemini 3.1 Flash Image, $1.50 on Gemini 3.1 Flash Lite Image and $12 on Gemini 3 Pro Image (half in Batch where Batch exists). How many thinking tokens an image uses depends on the prompt, so the estimate leaves them out and a Google image costs more than shown; the usage figures on a test call give the real count. Also excluded: reference images for editing (billed as input: Grok Imagine Image 2.0 charges $0.01 each, and OpenAI and Google bill input image tokens), the text model's tokens when you generate through OpenAI's Responses API, streamed partial previews (100 extra output tokens each on GPT Image 2.5), Google Search grounding, free tiers, subscription apps such as ChatGPT or Midjourney, regional uplifts and taxes.