Token counter for AI models
Paste a prompt, a system message, or a source file and get the token count for the model you use, plus characters, words, and how much of the context window it fills. OpenAI counts are exact for models with a published tokenizer mapping; GPT-6 counts are clearly labeled estimates until OpenAI publishes its mapping, and every other provider is an estimate as well. Your text is counted in memory and never stored or logged.
Your text is counted in memory on our server and is never stored, logged, or used for anything else. OpenAI counts are exact for models with a published tokenizer mapping. GPT-6 counts are estimates built on o200k_base until OpenAI publishes its tokenizer mapping, and are marked estimate in the model list. Counts for every other provider are estimates built on the closest public tokenizer (cl100k_base), because their own tokenizers are not available to this tool, and are marked estimate as well. Treat estimates as planning figures, not billing figures.