Vision Input Cost Estimator
Images cost tokens too. Drop an image to read its dimensions, or type width and height by hand, and see how many input tokens it uses on Claude and roughly what your monthly vision workload would cost.
Your image is never uploaded. It is read locally by your browser (FileReader) just to get its pixel dimensions - this page makes no network calls with your data.
Drop an image here, or click to pick one
PNG, JPG, GIF, WebP - only the width and height are read
:
![Local preview of the selected image]()
Or enter dimensions manually (no image needed)
0
Claude tokens / image
0
Tile-based estimate / image (approx.)
0
Claude image tokens / request
0
Claude image tokens / month
Monthly cost of the image tokens (Claude API prices)
| Model | Input $/1M | Cost / request | Cost / month |
|---|
How the estimate works: for Claude, if an image's long edge exceeds 1568 px the API first scales it down to fit (aspect ratio preserved), then charges roughly tokens = (width x height) / 750 - so a full 1568 px image tops out near 1,600 tokens no matter how large the original is. Anthropic documents this formula as the standard estimate. The tile-based number is a generic approximation of how some other providers (for example OpenAI-style tiling) bill images: a fixed base plus a fee per 512 px tile after their own downscaling rules - treat it as a ballpark only and check the provider's docs. Claude prices are the published Anthropic API input rates (as of July 2026): Fable 5 $10, Opus 5 / Opus 4.8 $5, Sonnet 5 / Sonnet 4.6 $3, Haiku 4.5 $1 per 1M input tokens. Costs shown cover the image tokens only; your text prompt and the model's output are extra - see the LLM Price Calculator.