If you’ve ever looked at AI pricing pages or API documentation, you’ve seen the word “token” everywhere. ChatGPT charges per token. Claude charges per token. Gemini, too. But what is a token, and why does it determine how much you pay?
Understanding tokens isn’t just technical trivia. It’s the key to controlling costs, writing better prompts, and knowing exactly what you’re paying for every time you hit “send.”
What Counts as a Token?
A token is a chunk of text that an AI model processes. It’s not quite a word, and it’s not quite a character—it’s somewhere in between.
In English, one token is roughly four characters, or about three-quarters of a word. The word “artificial” is two tokens. “AI” is one token. A short sentence like “How are you?” is four tokens (including punctuation and spaces).
Here’s where it gets interesting: tokens aren’t the same across languages. In languages like Chinese or Japanese, a single character can be multiple tokens. That means the same idea costs more to process in some languages than others.
Every major AI model—ChatGPT, Claude, Gemini—uses tokenization. OpenAI’s models use a system called tiktoken. Anthropic and Google use similar approaches. The exact token count varies slightly by model, but the concept is universal.
Why Tokens Determine What You Pay
AI companies charge by the token because that’s how much work the model does. When you send a prompt to ChatGPT, the model doesn’t just process your question—it processes every word you’ve sent in the conversation so far, plus every word it generates in response.
Let’s say you paste a 2,000-word article into Claude and ask for a summary. That’s roughly 2,600 input tokens. Claude generates a 200-word summary—about 270 output tokens. You just used 2,870 tokens in a single exchange.
Most AI companies charge separately for input tokens (what you send) and output tokens (what the model generates). Output tokens typically cost more—sometimes two to three times as much—because generating text requires more computation than reading it.
For example, as of September 2026, GPT-4o charges about $2.50 per million input tokens and $10 per million output tokens. That sounds cheap until you realize a single back-and-forth conversation with a long document can rack up tens of thousands of tokens.
How to See Your Token Usage
If you’re using ChatGPT, Claude, or Gemini through their web interfaces with a subscription, you don’t see token counts directly—you’re paying a flat monthly fee for unlimited use (with some rate limits). But if you hit those limits or use the API, tokens matter a lot.
API users can check token counts in the response data. OpenAI, Anthropic, and Google all return token usage in their API responses, showing you exactly how many input and output tokens each request consumed.
There are also free tools like OpenAI’s Tokenizer (platform.openai.com/tokenizer) that let you paste text and see how many tokens it uses. This is useful if you’re trying to estimate costs before running a large batch of prompts.
Practical Ways to Reduce Token Usage
If you’re paying per token or trying to stay within rate limits, here are a few ways to cut down:
- Start fresh conversations. Every message in a thread adds to the input token count. If you’re done with a topic, start a new chat instead of continuing a long thread.
- Trim unnecessary context. Don’t paste entire documents if a summary or key excerpt will do. The model doesn’t need your whole email chain—just the relevant parts.
- Use concise prompts. You don’t need to be polite or verbose. “Summarize this” works just as well as “Could you please provide a summary of the following text?”
- Limit output length. Ask for a 100-word summary instead of leaving it open-ended. The model will generate fewer tokens.
Why This Matters Beyond Cost
Tokens aren’t just about money. They also determine what fits in a model’s context window—the total amount of text it can process in one go. If your conversation uses too many tokens, the model starts “forgetting” earlier parts of the thread.
Understanding tokens helps you structure better prompts, manage long conversations, and avoid running into limits. Whether you’re using AI for work, research, or personal projects, knowing what a token is gives you more control over how the tool behaves—and what it costs.
Want one useful AI idea in your inbox every day? Subscribe to the One Two Three AI newsletter and stay ahead without the overwhelm.