Image Token Calculator
Count image input tokens and cost for vision models.
Vision models don't charge per image — they turn each one into tokens and bill those at the normal input rate, so a big photo can cost more than a page of text. This tool does the vision model image tokens math for you: enter the width and height or pick a preset, choose high or low detail, and see the tokens and gpt-4o image cost live. It follows OpenAI's documented image input token math for the GPT-4o family, so you can size a batch of images before you send them.
Your image
Tokens per image
765
1024×1024 at high detail
- Images × 1
- 765
- Cost per image
- $0.001912
- Total cost
- $0.001912
How high-detail tokens are counted
The image is scaled to fit 2048×2048, then down so its shortest side is 768px, then divided into 512px tiles. Tokens = 85 base + 170 per tile. Image tokens are billed at the model's normal input rate.
This follows OpenAI's documented tiling maths for the GPT-4o family, so counts for other providers (Claude, Gemini) or smaller models like GPT-4o mini will differ. Prices updated January 2026 — verify against the provider before you budget.
How it works
- 1
Set the image size
Enter the pixel width and height, or pick a preset like a 1080p screenshot or a phone photo. Typing a custom size switches off the preset.
- 2
Choose the detail level
High detail tiles the image and costs more; low detail is a flat 85 tokens regardless of size. Set how many images you're sending to price a batch.
- 3
Read tokens and cost
The headline is tokens per image. The stats show the batch token total, the cost per image and the total against the vision model you pick.
Instant & 100% private — nothing is uploaded
Every calculation runs locally in your browser. The prompts, token counts and numbers you enter stay on your own device and are never sent to a server — nothing is stored, logged or shared.
Frequently asked questions
- How are image tokens counted?
- For high detail, the GPT-4o family scales the image to fit 2048×2048, then scales it down so the shortest side is 768 pixels, divides it into 512-pixel tiles, and charges 85 base tokens plus 170 per tile. A 1024×1024 image works out to 765 tokens. Low detail skips the tiling and is a flat 85 tokens.
- How much does a GPT-4o image cost?
- Multiply the image's token count by the model's input rate. At 765 tokens for a 1024×1024 image and GPT-4o's input price, that's a fraction of a cent, but it adds up fast across a large batch or high-resolution photos with many tiles. The tool shows the exact figure for your size and count.
- Does this work for Claude or Gemini vision?
- The token math here is OpenAI's documented formula for the GPT-4o family. Anthropic and Google count image tokens differently, and smaller models like GPT-4o mini apply their own multiplier, so use this as a GPT-4o estimate and check other providers' docs for their method.
- Should I use high or low detail?
- Low detail's flat 85 tokens is much cheaper and fine when you only need the gist of an image — a rough layout or a yes/no check. High detail is worth the extra tokens when the model has to read small text, fine features or dense charts. Match the setting to what the task actually needs.
Important
For planning and estimates only. Prices come from a published rate table dated on the page; providers change pricing without notice, and token counts here are approximations. Confirm against the provider’s own pricing before you budget or commit.
More tools
More from the Hivly network
Free sister tools on our other sites.
