AI API cost calculator
Tokens per request, requests per month and model prices, as a monthly bill.
FreeRuns in your browserNo signupUpdated 2026-10-29
Rates last checked October 2026. Prices change; confirm on the provider's pricing page before you budget. Gemini 4 Argon is due to rise to $4 / $20 on 1 December 2026, and Gemini 3.8 Flash to $1.50 / $7.50 on 1 January 2027.
Cost per request
$0.0035
GPT-6.1 Terra · 43% of it is input
Per month
$70
$840 per year · 38M tokens
| Model | In / out per 1M | Per request | Per month |
|---|---|---|---|
| GPT-6 Astra | $10 / $50 | $0.035 | $700 |
| GPT-6.1 Sol | $2 / $10 | $0.007 | $140 |
| GPT-6.1 TerraSelected | $1 / $5 | $0.0035 | $70 |
| GPT-6 LunaCheapest | $0.1 / $0.5 | $0.0004 | $7 |
| Claude Fable 5.1 | $10 / $50 | $0.035 | $700 |
| Claude Opus 5.5 | $4 / $20 | $0.014 | $280 |
| Claude Sonnet 5.5 | $2 / $10 | $0.007 | $140 |
| Claude Haiku 5.5 | $0.1 / $0.5 | $0.0004 | $7 |
| Gemini 4 Argon | $2 / $10 | $0.007 | $140 |
| Gemini 3.8 Flash | $0.75 / $3.75 | $0.0026 | $52.5 |
| Gemini 3.5 Flash-Lite | $0.3 / $2.5 | $0.0015 | $29 |
Standard, uncached list prices. Claude Haiku 5.5 switches to $0.50 / $2.50 above 100K input tokens, and OpenAI's rates shown apply up to 272K. Prompt caching, batch APIs and volume discounts can cut the input cost a lot.
About this tool
How it works and how to read the result.
Need this done for you?
AI features for your product
Add AI search, chat and generation to your existing app.
See the serviceWhat it does
An OpenAI API cost calculator turns three numbers you can measure, input tokens per request, output tokens per request and requests per month, into a monthly and yearly bill. It covers the current GPT, Claude and Gemini models at their published list prices, so you can compare them for the same workload instead of reading nine pricing pages. The calculator runs in your browser and stores nothing.
How to use it
- Measure a typical request with the token counter: the system prompt, conversation history and user message together are the input; the reply is the output.
- Enter those two figures and your expected requests per month.
- Pick the model you plan to use. The comparison table updates for every model at once.
- Copy the comparison into your proposal or budget.
How to read the result
Take a chatbot that sends 1,500 input tokens and gets 400 output tokens back, 20,000 times a month. On GPT-6.1 Terra at $1 per million input and $5 per million output, one request costs $0.0035 and the month $70. The same traffic on Claude Opus 5.5 at $4 and $20 costs $0.014 per request, or $280 a month. On GPT-6 Luna or Claude Haiku 5.5, at $0.10 and $0.50, it drops to $7.
The share of cost that is input tells you where to save. If input is 70% or more, a shorter system prompt or prompt caching will help most. If output dominates, ask for shorter answers or structured output.
In this example the most capable models, GPT-6 Astra and Claude Fable 5.1, cost $700 a month, 100 times the cheapest. Long prompts and scheduled changes also move the figures: Claude Haiku 5.5 costs five times more above 100K input tokens, and Gemini 4 Argon doubles to $4 and $20 on 1 December 2026. Prototype on the capable one, then test whether a smaller model gives acceptable answers for your task before you scale.
Official pricing pages: OpenAI, Anthropic, Google Gemini. We budget model costs this way when scoping AI product features.
Questions
Frequently asked.
Where do the prices come from?
They are the published list prices per million tokens on each provider's pricing page, last checked October 2026. Providers change prices often, and new models arrive with different rates, so confirm on the official page before you commit a budget. See the links at the end of this page.
Why are output tokens so much more expensive?
Generating text is slower and uses more compute than reading it, so every provider charges several times more for output than input. Short, structured replies cost far less than long prose. If you only need a yes or no, ask for one.
Does this include caching and batch discounts?
No. The figures are for standard calls with no cached prompt. Prompt caching cuts the price of repeated input such as a system prompt, often by half or more, and batch APIs discount jobs that can wait. Both can change the picture if your volume is high.
More in AI & automation
Related tools.
Your next move
Like what you see? Tell us about it.
A process that takes too long, a website that undersells you, or an idea ready to take shape. Tell us where you are and we will suggest a useful next step.
India · Working with clients worldwide