Vision Token Cost Calculator
Calculate the token cost of sending images to a multimodal vision-language model.
About Vision Token Cost Calculator
Vision-enabled LLMs like GPT-4V convert images into a fixed or resolution-dependent number of tokens before billing. Low-detail mode charges a flat token count (commonly 85 tokens) regardless of image size, while high-detail mode tiles the image into 512×512 segments and charges roughly 170 tokens per tile plus a small base. This calculator estimates total token usage and cost for a batch of images under either mode.
Formulas
Tokens per image (high detail)
tokens_per_image = 85 + tiles_per_image × 170
- 85
- Base token overhead
- 170
- Tokens per tile
- tiles_per_image
- Number of 512×512 tiles covering the image
Total cost
total_cost = images × tokens_per_image × (rate_per_1M_tokens / 1000000)
- images
- Number of images
- tokens_per_image
- Tokens consumed per image
- rate_per_1M_tokens
- Dollar cost per million tokens
Worked example
With the default values below, Vision Token Cost Calculator returns Total Cost of $0.0043. Change any input above and every figure here is recalculated in your browser.
Inputs used
- Number of Imagesaccepts 1–1,000,000 images
- 10 images
- Detail Levelone of Low Detail (fixed 85 tokens), High Detail (tiled, resolution-based)
- Low Detail (fixed 85 tokens)
- Tiles per Image (high detail only)accepts 1–64
- 6
- Rate per 1M Tokensaccepts 0–200 $/1M tokens
- 5 $/1M tokens
Results
- Total Cost
- $0.0043
- Total Tokens
- 850 tokens
- Tokens per Image
- 85 tokens
Step by step
- 1Tokens per image (low detail, fixed)= 85 tokens
- 2Total tokens= 850 tokens
- 3Total cost= $0.0043
Assumptions and limits
- The result is derived only from the 4 inputs listed above. Anything not among them is not part of the calculation.
- In particular, this calculator does not take inflation, tax or compounding frequency into account.
- Figures are rounded for display to 4 decimal places, so totals built from several rounded parts can differ slightly from the unrounded arithmetic.
How these calculators are built and tested is described on the Methodology page. If a figure looks wrong, please report it.
Frequently Asked Questions
When should I use low detail mode?
Use low detail for thumbnails, icons, or cases where fine text/detail recognition isn't needed — it's dramatically cheaper per image.
How many tiles does a typical image need?
A 1024×1024 image typically needs about 4 tiles; larger or non-square images may need more. Check your provider's tiling documentation for exact counts.
Does this apply to all vision models?
The tile/token math shown mirrors OpenAI's GPT-4V-style approach; other providers (Claude, Gemini) use different image-to-token conversion schemes.