If you’ve used ChatGPT, Claude, or any other AI chatbot, you’ve probably seen messages about “token limits” or noticed pricing based on “tokens per million.” But what exactly is a token, and why should you care?
Understanding tokens isn’t just technical trivia—it directly affects how much you pay, whether your prompts will work, and how to get better results from AI tools. Let’s break it down in plain English.
What Is a Token?
A token is the basic unit AI models use to process text. Think of tokens as chunks of text—but they’re not quite words and not quite characters.
Here’s how it typically breaks down:
- One token ≈ 4 characters in English
- One token ≈ ¾ of a word on average
- 100 tokens ≈ 75 words
- 1,000 tokens ≈ 750 words (about one page of text)
For example, the sentence “AI tokens determine pricing” contains 5 words but uses about 6-7 tokens. Common words like “the” or “is” are usually one token, while longer or unusual words might be split into multiple tokens. The word “understanding” might be two tokens: “under” and “standing.”
Punctuation counts too—a comma or period is typically its own token. Spaces are included in tokens with the words that follow them.
Why Tokens Matter for Your Wallet
Every AI service charges based on tokens, not words or messages. Both your input (what you send) and output (what the AI responds with) count toward your usage.
Current pricing examples (as of June 2024):
- ChatGPT Plus: $20/month for unlimited messages, but GPT-4 has context limits
- Claude Pro: $20/month with usage limits based on tokens
- API access: GPT-4 costs about $10 per million input tokens, $30 per million output tokens
- GPT-3.5 API: Much cheaper at around $0.50 per million input tokens
If you’re pasting entire documents into ChatGPT, you could be using thousands of tokens in a single conversation. A 10-page document might be 7,500 tokens—and that’s before the AI’s response.
How Token Limits Affect What You Can Do
Every AI model has a context window—the maximum number of tokens it can handle at once, including your conversation history and its response.
Common context windows:
- GPT-4: 8,000 or 32,000 tokens (depending on version)
- GPT-4 Turbo: 128,000 tokens
- Claude 3: 200,000 tokens
- Gemini 1.5 Pro: Up to 1 million tokens
When you hit the limit, the AI starts “forgetting” earlier parts of your conversation. This is why ChatGPT sometimes loses track of instructions you gave at the start of a long chat.
If you’re trying to analyze a long document, compare multiple files, or have an extended back-and-forth conversation, you need to know these limits. Upload a 50-page PDF to a model with an 8,000-token limit, and it simply won’t work.
Practical Tips for Managing Tokens
Check your token count. Use tools like OpenAI’s Tokenizer (platform.openai.com/tokenizer) to see exactly how many tokens your text contains before you send it. This helps you stay under limits and estimate costs.
Be concise with prompts. Every word in your prompt counts against your limit and costs money. Get to the point. Instead of “I would really appreciate it if you could help me understand,” just write “Explain.”
Summarize long conversations. When a chat gets long, ask the AI to summarize the key points, then start a fresh conversation with that summary. This resets your token count while preserving important context.
Use the right model for the job. Don’t use GPT-4 when GPT-3.5 will do. Simple tasks like formatting text, basic summaries, or grammar checks work fine on cheaper models with smaller token costs.
Split large documents. If you need to process a very long document, break it into sections and process each separately, then combine the results.
Choose models based on context needs. If you’re working with long documents regularly, Claude or Gemini’s larger context windows might save you time and frustration, even if per-token costs are similar.
The Bottom Line
Tokens are how AI models read, process, and charge for text. One token is roughly three-quarters of a word, and both your questions and the AI’s answers count toward usage limits and costs.
Knowing this helps you pick the right tool, stay under limits, and avoid surprise charges. Check your token counts for big jobs, keep prompts focused, and match your model choice to your actual needs.
You don’t need to obsess over every token, but understanding the basics puts you in control of your AI spending and helps you work more effectively.
Want more practical AI tips like this? Subscribe to the One Two Three AI newsletter and get one useful AI idea delivered to your inbox every day—no hype, just helpful advice you can actually use.