Paste any prompt to count its tokens the way OpenAI’s tokenizer does, then see what it costs on 114 AI models from 12 labs, and what the same job would have cost in 2023.
Loading tokenizer…
OpenAI counts are exact for models that use these encodings. Anthropic and Google don’t publish browser tokenizers, so their figure is a range based on the OpenAI count. Anthropic’s newer tokenizers (Claude Opus 4.7 onward) can produce about 30% more tokens for the same text. Your text never leaves your browser.
AI API cost calculator
Estimate cost per request and per month. Input tokens fill in from your text automatically, or type any number.
| Model | Lab | Input $/1M | Output $/1M | Per request | Per month |
|---|
List prices per 1M tokens for standard, uncached requests, from The Brief Script’s AI API Price Index. Batch, cached-input and long-context tiers change real bills. Check each provider’s pricing page before you commit.
The same job in 2023 vs today
Your workload, priced on the best model of March 2023 and on today’s cheapest model that matches it.
GPT-4-level output for this job is — cheaper than in 2023. The best models still cost real money.
“GPT-4-level” means at or above GPT-4’s capability score in our price index. See the methodology.
What is a token?
A token is the unit an AI model reads and writes. Tokenizers split text into common chunks: a short word is often one token, a long or rare word becomes several, and spaces and punctuation count too. Every AI API bills by the token, so the count above is what you pay for.
Tokens to words: a quick guide
OpenAI’s rule of thumb for English text is that one token is roughly four characters, or three quarters of a word. Use the counter for exact figures; this table is for quick planning.
| Tokens | Approx. English words | Roughly equals |
|---|---|---|
| 1 | 0.75 | about 4 characters |
| 100 | 75 | a short paragraph |
| 1,000 | 750 | a long email or a page and a half |
| 10,000 | 7,500 | a long report |
| 128,000 | 96,000 | a full-length book |
| 1,000,000 | 750,000 | about 10 novels |
How this token counter works
The tool loads OpenAI’s published byte-pair encodings, o200k_base and cl100k_base, through the open-source js-tiktoken library and runs them in your browser. o200k_base is the encoding OpenAI introduced with GPT-4o; cl100k_base covers GPT-4 and GPT-3.5. The colored chips show exactly where each token starts and ends, which helps when you’re trimming a prompt.
Claude and Gemini use their own tokenizers. Their counts are usually close to OpenAI’s, but not identical, so we show a range. For exact numbers, both Anthropic and Google offer token-counting endpoints in their APIs.
How to estimate your AI API bill
Cost per request = (input tokens × input price + output tokens × output price) ÷ 1,000,000. Multiply by requests per day and by 30 for a monthly figure. The calculator prices all 114 models in our index at once, so you can see the gap between a flagship and a budget model for your exact workload.
Five ways to cut token costs
- Right-size the model. Routine extraction and classification rarely need a flagship. The table above often shows a 50x or bigger gap for the same job.
- Cap output length. Output tokens usually cost several times more than input tokens.
- Use prompt caching. Most major labs discount repeated input, such as long system prompts, by up to about 90%.
- Batch non-urgent work. OpenAI, Anthropic and Google offer batch APIs at roughly half price for jobs that can wait.
- Trim the prompt. Paste it into the counter, remove boilerplate, and watch the count fall.
Where the prices come from
Prices come from The Brief Script’s AI API Price Index, which tracks list prices, price changes and retirements for 197 models across 12 labs since March 2023. The calculator shows the 114 models still on sale as of Oct 29, 2026, at standard, uncached rates. We update it monthly. Want to know which assistant fits your work? Read our Gemini vs ChatGPT vs Claude scorecard.
Frequently asked questions
How many tokens is 1,000 words?
In English, about 1,300 tokens with OpenAI’s tokenizers. OpenAI’s rule of thumb is that one token is about four characters, or three quarters of a word. Code, numbers and non-English text use more tokens per word.
Is this token counter accurate for ChatGPT and GPT models?
Yes. It runs OpenAI’s own o200k_base and cl100k_base encodings in your browser through the open-source js-tiktoken library, so counts match what the OpenAI API uses for models built on those encodings.
Can I count tokens for Claude or Gemini here?
Only as an estimate. Anthropic and Google use their own tokenizers and don’t publish them for browsers, so we show a range based on the OpenAI count. For exact numbers, use each provider’s token-counting API.
How do I calculate the cost of an API request?
Multiply input tokens by the input price and output tokens by the output price, then divide by one million, because prices are quoted per million tokens. The calculator does this for every model at once and scales it to a month.
Why are output tokens more expensive than input tokens?
Generating text is slower and uses more compute than reading it, so most labs charge three to eight times more per output token. Keeping answers short is one of the easiest ways to cut a bill.
Is my text stored or sent anywhere?
No. Tokenizing and cost math run entirely in your browser. The page only downloads the public tokenizer files; your text never leaves your device.