Why image size matters
Patch-based models count the resized image, not always the original dimensions. Fixed-resolution budgets, such as Gemini 3 high, do not scale proportionally with pixel area.
Vision token estimates
Image Token Calculator
Upload an image or enter its dimensions to estimate how many tokens it costs on GPT-6, Claude, and Gemini — plus what a single image costs at current API prices.
Image inputs
Common sizes
Image size
1.0 MP
1,024 x 1,024 px
Cheapest per image
$0.0002
on GPT-5.6 Luna
Most tokens
1,369 tokens
on Claude Sonnet 5
Tokens and cost by model
Estimates from each provider's documented formula and current input pricing.
| Model | Tokens | Est. cost / image |
|---|---|---|
| GPT-6 AstraHigh detail: 32px patches, 2,500-patch budget, x 1.2OpenAI | 1,229 | $0.012 |
| GPT-5.6 SolHigh detail: 32px patches, 2,500-patch budget, x 1.2OpenAI | 1,229 | $0.0049 |
| GPT-5.6 LunaHigh detail: 32px patches, 2,500-patch budget, x 1.2OpenAI | 1,229 | $0.0002 |
| Claude Sonnet 528px patches after resizing; high-resolution tier (4,784 tokens)Anthropic | 1,369 | $0.0027 |
| Claude Haiku 4.528px patches after resizing; standard tier (1,568 tokens)Anthropic | 1,369 | $0.0014 |
| Gemini 3.8 FlashGemini 3 high media resolution: approximately 1,120 tokensGoogle | 1,120 | $0.0008 |
| Gemini 3.5 Flash-LiteGemini 3 high media resolution: approximately 1,120 tokensGoogle | 1,120 | $0.0003 |
Resize before sending
Keep enough detail for the task. Small text, charts, and dense screenshots may need higher resolution; check legibility before reducing dimensions or choosing a cheaper mode.
Batch budgets
Multiply the per-image cost by your expected volume. A thousand phone photos can cost several dollars on premium models, but cents on lite ones.
Counting rules: OpenAI vision, Claude vision, and Gemini media resolution. Estimates cover image input only at standard per-token rates, excluding prompt text, output, caching, and long-context surcharges.
Image tokens FAQ
How many tokens is an image?
There is no universal image token count. The selected model's resize rules and detail or media-resolution setting determine the result. This calculator labels its assumed mode for each model and estimates image input only.
How does OpenAI count image tokens?
The OpenAI models listed here use 32px patches with a 1.2 multiplier. We calculate high detail with a 2,500-patch budget after resizing. Model-specific edge limits apply; automatic or original detail can use different budgets.
How does Claude count image tokens?
Claude counts 28px patches after resizing. Haiku 4.5 uses the standard tier; Sonnet 5 uses a higher-resolution tier. The calculator applies the appropriate edge and token limits before counting, rather than billing every pixel in the original image.
How does Gemini count image tokens?
For the Gemini 3 models listed here, this calculator assumes high media resolution: approximately 1,120 tokens per image. Other resolution settings differ. Older Gemini tiling rules should not be applied to this mode.
How can I reduce image token cost?
Choose a resolution that preserves the details your task needs. Lower detail modes may save tokens, and model input prices also matter. Reducing dimensions does not always reduce cost when a fixed media-resolution budget applies.
Is my uploaded image sent to a server?
No. This calculator reads only the image width and height locally in your browser. The image file itself is never uploaded anywhere.