Free tool

What will AI cost you each month?

Describe the work once and compare Claude, GPT and Gemini side by side, using each vendor's published prices. It runs in your browser; nothing you type is sent anywhere.

Prices checked

12 October 2026, from the official pricing pages of Anthropic, OpenAI and Google.

10 current models, standard tier, US dollars. Estimates, not quotes.

Start from an example

10,000 questions a month, instructions plus a few help articles per question

≈ 4,500 English words. Count instructions, history and any documents you include.

≈ 300 words

The repeated part of every prompt, such as fixed instructions, can be cached and billed at a fraction of the price.

Estimated monthly cost

Cheapest here: Claude Haiku 5.5 at $6.38 a month ($0.0006 per request). The cheapest model is not always the right one: small models answer simple, well-defined questions well and struggle with long reasoning. Test on your own examples before you commit.

  • Claude Haiku 5.5Anthropic · Small & fast
    $6.38
  • GPT-6 LunaOpenAI · Small & fast
    $6.38
  • Gemini 3.5 Flash-LiteGoogle · Small & fast
    $23.14
  • Gemini 3.8 FlashGoogle · Mid-range
    $47.85
  • Claude Sonnet 5.5Anthropic · Mid-range
    $126
  • GPT-6.1 SolOpenAI · Mid-range
    $126
  • Gemini 3.1 Pro (preview)Google · Flagship
    $136
  • Claude Opus 5.5Anthropic · Flagship
    $252
  • Claude Fable 5.1Anthropic · Flagship
    $625
  • GPT-6 AstraOpenAI · Flagship
    $638

Want us to check your real numbers?

Free 15-minute AI cost review. Send us these figures and what the AI does, and we'll reply with the cheapest setup that still answers well: model, caching, batch and prompt size.

Get the free cost review →
How to read it

The model price is rarely the biggest number.

What you send decides the bill more than which model you pick. A retrieval system that finds the right three passages costs far less than one that pastes in the whole manual, and usually answers better too.

What is a token?
Models read and write text in small pieces called tokens. In English, 1,000 tokens is roughly 750 words. Each vendor counts slightly differently, so treat the word estimates here as approximate.
Why do long prompts cost more than the numbers suggest?
Some models switch to a higher price once a single request's prompt passes a size limit, and then the whole request is billed at that rate. Claude Haiku 5.5 does this above 100,000 tokens, Gemini 3.1 Pro above 200,000 and OpenAI's models above 272,000. The calculator flags it when your prompt crosses the line.
Is the same text the same number of tokens on every model?
No. Anthropic says its newer models (Claude 4.7 and later) produce about 30% more tokens for the same text than its older ones. Different vendors' tokenizers also differ. For a precise figure, count a few real requests with each vendor's token counter.
What do caching and batch do?
Caching bills the repeated part of a prompt, such as fixed instructions, at a small fraction of the normal input price. Batch means sending work that can wait hours for an answer; all three vendors charge half price for it.
Is my data sent anywhere?
No. The calculator runs entirely in your browser and nothing you type leaves this page.

Want this running in your business?

Tell us what you want AI to do. We'll tell you which model fits, what it will really cost, and build it.

Start a project