AI API price calculator: text, image, video and 3D
Price a whole workload on several platforms side by side. Tokens in and out, image size and quality, video resolution, length and audio, 3D texture and PBR all change the price, so each job is priced the way each platform bills it.
A job is a batch of images, priced by size (1K to 4K, or your own) and quality.
Cheapest for these jobs: APIMart Ext at $7.00. The most expensive exact total is Official at $50.80, 7.3× as much. The cheapest is 86% below the official price, $50.80.
| Job | ||||||
|---|---|---|---|---|---|---|
images | ||||||
images | ||||||
images | ||||||
| Total | $50.807.3× the cheapest | −86%$7.00 | −41%$30.004.3× the cheapest | −8%$46.806.7× the cheapest | −7%$47.416.8× the cheapest | −2%$50.007.1× the cheapest |
- Cheapest exact price
- −38% +16%Against the official price for the same job (shaded column)
- ≈ Approximate match: another size, quality or length, no output or cache price, or far above the official price
- Estimate: worked out from a token rate
- $Plan or starting price
- Grey note: the platform doesn’t say (audio, texture), so it is one price either way and still compared
How each job is priced
Each cell is what the model’s compare page quotes for that job, from the same price ledger, times the count. A text request is its fresh input at the input price, its cached share at the cache-read price (the input price where none is sold) and its output at the output price; a prompt past a platform’s band ("prompt > 200K") is priced at that band, and models with Batch or Thinking tiers let you pick the billing mode per job. A price that doesn’t say with or without audio, or textured or not, is taken as one price either way. A 3D job wanting PBR maps, on a platform whose price doesn’t include them, is shown greyed at the textured price; low-poly and other add-ons aren’t priced. Per-megapixel prices are charged for the job’s pixels (2K is 2048 × 2048; a custom size is its own width × height). On each platform, a custom size is priced at the smallest size that platform sells with as many pixels, within 10% (1920 × 1080 is 1.5K where 1.5K is sold, else 2K). GPT Image and Nano Banana listings billed only per output image token are estimated with the maker’s official token counts. EvoLink’s per-clip prices come from one per-second rate, so a length between two listed ones is priced at that rate. Seedance billed in video tokens is estimated with the formula fal publishes: width × height × seconds × 24 / 1024. On credit-plan "from" platforms, turning on plan/promo rates uses their yearly plan, credit-pack or partner promo discount while still keeping those prices marked as non-exact. Top-up fees, taxes and OpenRouter’s credit fee are not included.