AI Token Counter
Paste any text below to instantly estimate how many tokens it uses for GPT, Claude and Gemini, see whether it fits each model's context window, and price a single API call across every major model. It all runs in your browser, so nothing you paste is uploaded or saved.
The fast answer
- 1 token is about 4 characters of English, so roughly 750 words is about 1,000 tokens.
- Input and output are billed separately, and output usually costs 2 to 6 times more, so a long answer can dwarf the cost of your prompt.
- This is an estimate. Each model tokenizes a little differently (usually within 10 to 20 percent); for billing-critical numbers, use the provider's own tokenizer.
Different tokenizers split the same text slightly differently. These are approximations, not exact counts.
Context window fit
A chip turns green when your text fits with at least 25 percent of the window left free for the model's reply.
| Provider | Model | Input $/1M | This text |
|---|---|---|---|
| DeepSeek | DeepSeek V4 Flash | $0.14 | $0.00 |
| Gemini 3.1 Flash-Lite | $0.25 | $0.00 | |
| OpenAI | GPT-5 mini | $0.25 | $0.00 |
| Anthropic | Claude Haiku 4.5 | $1.00 | $0.00 |
| Anthropic | Claude Sonnet 4.6 | $3.00 | $0.00 |
| OpenAI | GPT-5.5 | $5.00 | $0.00 |
| Anthropic | Claude Opus 4.8 | $5.00 | $0.00 |
Input-only cost for sending the pasted text one time, at standard 2026 list prices. The model's reply is billed separately at the output rate (usually higher). Want your real monthly bill across input and output volume? Use the LLM API pricing estimator.
Questions about tokens
What is a token in AI models?
A token is the unit large language models read and bill on. It is a chunk of text, usually a common word, part of a longer word, or a piece of punctuation. In English, one token is roughly 4 characters, so about 750 words works out to roughly 1,000 tokens. Every API charges per token, split into input tokens (what you send) and output tokens (what the model writes back).
How accurate is this token counter?
It is a fast estimate, not an exact count. Each model family uses its own tokenizer, so GPT, Claude and Gemini can each split the same text a little differently, typically within 10 to 20 percent of each other. This tool uses a character and word based approximation that is very close for normal English prose but can drift on code, non-English text, or long strings of numbers. For billing-critical numbers, run the provider's official tokenizer.
Will my text fit in the context window?
The context window is the maximum number of tokens a model can hold at once, including your prompt and its answer. Common 2026 windows are 128K for many open models, 200K for Claude Haiku, 400K for GPT-5 class models, and 1,000,000 for Claude Sonnet, Claude Opus, and Gemini. The chips above show whether your text fits, with room left for the model's reply.
How do I lower my token cost?
Send less input and ask for shorter output. Trim system prompts, drop irrelevant context, and summarize long documents before sending them. Because output tokens cost 2 to 6 times more than input, telling the model to be concise saves the most. For repeated context, prompt caching (offered by Anthropic, OpenAI and Google) can cut input cost by up to 90 percent. To compare models by your real monthly volume, use our LLM API pricing estimator.
Is my text private?
Yes. The counting runs entirely in your browser. Nothing you paste is uploaded, logged, or sent to any server, so you can safely check prompts that contain sensitive or unpublished text.
Now find the right model for the job
Token count is only half the decision. See which model wins on reasoning, coding, speed and real-world cost before you commit.
ChatGPT vs Claude vs Gemini → · LLM API pricing → · Compare all tools →