AI cost calculator for ChatGPT, Claude and Gemini
Enter how much you use AI and see what it costs per request, per month and per year on every current model from OpenAI, Anthropic and Google. Then see how much shorter prompts and answers would save.
Instructions and chat history ride along with every reply.
Monthly cost on every major model
Cheapest first, at today's pay-per-use list prices. Claude rows include its higher token count for the same text.
| Model | Level | Per request | Per month | Per year |
|---|---|---|---|---|
| GPT-6 LunaOpenAI | Light | $0.00020 | $6.04 | $72.54 |
| Claude Haiku 5.5Anthropic | Light | $0.00027 | $8.16 | $97.93 |
| Gemini 3.1 Flash-LiteGoogle | Light | $0.00055 | $16.58 | $199 |
| Gemini 3.8 FlashGoogle | Everyday | $0.0015 | $45.34 | $544 |
| GPT-6.1 TerraOpenAI | Everyday | $0.0020 | $60.45 | $725 |
| GPT-6.1 SolOpenAI | Hard reasoning | $0.0040 | $121 | $1,451 |
| Gemini 4 ArgonGoogle | Hard reasoning | $0.0040 | $121 | $1,451 |
| Claude Sonnet 5.5Anthropic | Everyday | $0.0054 | $163 | $1,959 |
| Claude Opus 5.5Anthropic | Hard reasoning | $0.011 | $326 | $3,917 |
| GPT-6 AstraOpenAI | Frontier | $0.020 | $605 | $7,254 |
| Claude Fable 5.1Anthropic | Frontier | $0.027 | $816 | $9,793 |
These are estimates: 1,040 tokens sent and 195 tokens back per request on ChatGPT and Gemini, about 35% more on Claude. Check real text with the token counter, and see today's prices for every model.
How the calculator works
Every major AI API charges per token, with separate prices for the text you send (input) and the text the model writes (output). The calculator converts your words to tokens, at about 1.3 tokens per English word on ChatGPT and Gemini and roughly 35% more on Claude, then applies each model's published list price.
- Words sent per request should include everything that goes with it: your instructions, any pasted document, and the chat history, which is re-sent with every message.
- Words in each answer should include hidden reasoning if you use a thinking mode, since it's billed as output even though you never see it.
- Prices change on schedule. When a company announces a new price, the calculator shows what your workload will cost from that date.
These are list prices. Prompt caching and batch processing can cut them further; our AI pricing comparison covers the discounts each company offers, and how to reduce AI costs covers what to change first.
Where the savings are
Move the sliders and you'll see the pattern: for most workloads, choosing a smaller model saves far more than anything else, and shorter answers save more than shorter prompts. The prompt optimizer does both for a single prompt: it rewrites it shorter and names the cheapest model at each company that handles it well.
Frequently asked questions
How do I calculate AI API costs?
Multiply the tokens you send by the model's input price, add the tokens it writes times its output price, and multiply by how many requests you make. Prices are quoted per million tokens, and 100 tokens is roughly 75 English words on ChatGPT and Gemini. This calculator does the math for every current model at once.
How much does the OpenAI (ChatGPT) API cost?
OpenAI's current models cost from $0.10 per million input tokens and $0.50 per million output tokens (GPT-6 Luna) up to $10 and $50 (GPT-6 Astra), at list prices checked October 30, 2026.
How much does the Claude API cost?
Anthropic's current Claude models cost from $0.10 per million input tokens and $0.50 per million output tokens (Claude Haiku 5.5) up to $10 and $50 (Claude Fable 5.1). Claude also counts roughly 35% more tokens than ChatGPT for the same English text, which this calculator includes.
How much does the Gemini API cost?
Google's current Gemini models cost from $0.25 per million input tokens and $1.50 per million output tokens (Gemini 3.1 Flash-Lite) up to $2 and $10 (Gemini 4 Argon). Google has announced price increases for some models; the calculator shows what your workload would cost after each one.
Is this the same as a ChatGPT Plus or Claude Pro subscription?
No. Chat subscriptions charge a flat monthly fee with usage limits. This calculator is for pay-per-use API pricing, which is how businesses and apps are billed when they build on these models.
What's the fastest way to lower an AI bill?
Use the smallest model that does the job well, cap answer length, and stop re-sending text the model doesn't need. Answers usually cost about five times more per token than prompts, so shorter answers save the most. Our guide to cutting AI costs covers caching and batch discounts too.