AI Token Cost Calculator
Work out exactly what an AI model will cost per request, per day, per month and per year.
Calculate your API cost
Your estimated cost
Compare every model at your usage
Same tokens and same request volume, priced on each model. Your selected model is highlighted.
How AI token pricing works
AI APIs do not charge per question or per word. They charge per token, and tokens are small pieces of text, roughly three quarters of a word in English. Every request has two kinds of tokens: input tokens, which are everything you send (your prompt, any documents, and the conversation history), and output tokens, which are everything the model writes back. Output tokens almost always cost more than input tokens, usually three to five times more, because generating text is the expensive part.
The cost formula
Cost per request = (input tokens / 1,000,000 × input price) + (output tokens / 1,000,000 × output price). Multiply by requests per day for the daily cost, by 30 for a monthly estimate, and by 365 for a yearly estimate. This calculator uses exactly that formula.
Worked example
Say a support chatbot sends 1,500 input tokens and gets 500 output tokens per conversation, and handles 1,000 conversations a day. On GPT-5 at an approximate $1.25 per million input tokens and $10 per million output tokens, one conversation costs about $0.0069, which is $6.88 a day, about $206 a month and about $2,510 a year. On GPT-5 mini the same workload costs a small fraction of that, which is why choosing the right model size matters more than almost anything else in an AI budget.
Why the model choice changes everything
The price gap between the biggest flagship model and the smallest fast model from the same provider can be 25 times or more. Most everyday tasks, such as summarising, sorting, tagging and drafting, work perfectly well on the smaller models. A common pattern is to send simple requests to a small model and only escalate to a flagship model when the small one fails a quality check. The comparison table above shows what that decision is worth at your own numbers.
Four ways to cut your AI bill
- Trim the input. Long system prompts and pasted documents are charged on every single request. Shorten them, or summarise them once and reuse the summary. Not sure how big your prompt is? Count it first with our AI Token Counter.
- Cap the output. Set a maximum output length so the model cannot write a 2,000-token answer to a 50-token question.
- Use the smallest model that passes. Test a mini or Flash class model on real examples from your workload before paying flagship prices.
- Watch the conversation history. Chat apps resend the whole history each turn, so a long chat quietly multiplies input tokens. Summarise older turns.
Approximate prices used in this calculator
Input and output prices per 1 million tokens, in US dollars: GPT-5 $1.25 / $10.00; GPT-5 mini $0.25 / $2.00; GPT-5 nano $0.05 / $0.40; Claude Sonnet 4.5 $3.00 / $15.00; Claude Opus 4.1 $15.00 / $75.00; Gemini 2.5 Pro $1.25 / $10.00; Gemini 2.5 Flash $0.30 / $2.50; DeepSeek V3 $0.27 / $1.10. These are approximations of public list prices and change often; some providers also charge less for cached input tokens or batch processing, which this calculator does not model.
Frequently asked questions
How much does 1 million tokens cost?
It depends entirely on the model. One million input tokens costs roughly $0.05 on the smallest models and up to $15 on flagship models, and output tokens cost more than input tokens on every major model. Pick a model in the calculator and set requests to see the exact figure at your usage.
Why do output tokens cost more than input tokens?
Reading your prompt is one pass of computation, while writing an answer means generating text one token at a time, which takes far more computing power. Providers price output higher to reflect that, usually three to five times the input price.
How do I estimate tokens before I build anything?
As a rough guide, 1,000 tokens is about 750 words of English text. For a real prompt, paste it into our AI Token Counter to get an estimated token count plus character and word counts.
Are these the exact prices I will pay?
No. Prices here are approximate public list prices for comparison, and providers change them regularly. Your bill can also differ because of cached-token discounts, batch discounts, free tiers and regional pricing. Always confirm on the provider’s pricing page.
What is the cheapest way to run a chatbot?
Use a small fast model such as a mini, nano or Flash class model, keep prompts short, cap the answer length, and summarise long conversations instead of resending the full history. At high volume those four choices matter more than the provider you pick.
Does this calculator store or send my data anywhere?
No. The calculator runs entirely in your browser. Nothing you type is uploaded, stored or shared.
How are monthly and yearly costs worked out?
Monthly cost is the daily cost multiplied by 30 and yearly cost is the daily cost multiplied by 365, assuming the same request volume every day. If your traffic is seasonal or weekday-only, adjust the requests-per-day figure to match a typical day.
More tools you may need
Counting the tokens in a single prompt? Try the AI Token Counter. Running a website? Our SPF Record Generator and DMARC Record Generator build your email DNS records for you, and our guide explains how much AI tokens cost in plain English.
