Free developer tool
AI Token Calculator
Calculate the token count for your content based on the selected AI model.
Live count
Calculate the token count for your content
Review token, character, and word counts before sending your content to the model.
Your content
Enter content to analyze
Choose a model, then paste text or load a text-based file. Nothing is uploaded or saved.
Supported: .txt, .md, .json, .csv, and source-code files
Counts are calculated locally and discarded when you leave or refresh this page.
Know before you send
Understand your content size in tokens
Token counts help you understand the size of your prompt and source material before a request is made.
01
Local by default
Text is processed in memory by this page and is not sent to model providers or saved by EidoStack.
02
Model-aware results
Switch models to compare their token counts before choosing where a prompt should run.
Common Questions
Why do different AI models count tokens differently?
Different AI models may use different tokenizers and vocabularies. The same text can therefore produce a different number of tokens depending on the model or provider being used.
Are spaces and punctuation counted as tokens?
Spaces and punctuation can affect tokenization, but they are not always counted as separate tokens. A tokenizer may combine parts of words, spaces, punctuation marks, or symbols into individual tokens depending on its vocabulary.
Does code use more tokens than regular text?
It can. Source code contains punctuation, indentation, operators, variable names, and other structures that may be split into many tokens. The exact token count depends on the programming language, formatting, and tokenizer.
Do different languages use different numbers of tokens?
Yes. Token efficiency varies between languages because tokenizers represent words and characters differently. The same sentence translated into another language may require noticeably more or fewer tokens.
Why is token count important for AI prompts?
Token count determines how much text is sent to the model and how much of the model's context window is being used. Tracking prompt size helps prevent context-limit errors and makes it easier to design prompts that leave enough room for the model's response.
What happens if my prompt exceeds the model's context window?
If the combined input and generated output exceed a model's context limit, the request may fail, older context may need to be removed, or the application may need to shorten the input before sending it to the model.
Should I count the system prompt as part of the context?
Yes. System instructions are normally sent to the model together with the user prompt and other conversation context. They therefore contribute to the total number of input tokens used by the request.
Does conversation history increase token usage?
Yes. If previous messages are included with every new request, they become part of the model input. Long conversations can therefore consume an increasingly large portion of the available context window.
Can I count tokens in code, JSON, or structured data?
Yes. Token counters can be useful for source code, JSON, XML, Markdown, logs, and other structured content in addition to normal prose. This is especially useful when preparing large payloads for AI APIs.
Why can an estimated token count differ from the final API usage?
The final API token count may include additional content that is not visible in the text you manually count, such as system instructions, message formatting, tool definitions, retrieved context, or provider-specific formatting. For exact billing and usage data, the provider's API response should be treated as the final source of truth.