AI Token Counter & Cost Calculator

The unit language models read and bill by. A token is a chunk of text, usually a word or part of a word plus the space before it. In English, one token is roughly four characters or three-quarters of a word, so 1,000 tokens is about 750 words. Code, numbers, and non-Latin scripts cost more tokens per character.

Each vendor has its own tokenizer. OpenAI publishes its encodings through tiktoken, so the o200k_base count is exact for that encoding (tiktoken maps it to the GPT-5 family and lists no GPT-6 entry yet, so we treat it as the best available count for GPT-6). Anthropic and Google offer count-tokens API endpoints instead of a downloadable tokenizer; Anthropic's docs say Claude 4.7 and later produce about 30% more tokens than the older rule of 3.5 English characters per token, and Google puts a Gemini token at about four characters, so those rows are labeled estimates.

No. The tokenizer and the price table are downloaded to your browser once, and all counting and arithmetic happen there. There is no server endpoint behind this page, no account, and no analytics on what you paste.

No. Subscriptions charge a flat monthly fee with usage limits rather than per-token billing. The costs here are the vendors' standard API list prices, which is what you pay when you build on the model or use a tool that bills API usage through to you. Our AI app pricing index covers the subscription side.

Every model has a maximum number of tokens it can hold at once. The percentage shows how much of that limit your request would fill. Above about 60–70% you leave little room for conversation history and tool results; over 100% the request is rejected or truncated. Where the vendor defines the window as input plus output (OpenAI, Anthropic, Gemini 4 Argon) the figure includes your expected output tokens; where it publishes separate input and output budgets (Gemini 3.8 Flash) or input only (Grok, Muse), it counts your input alone.

OpenAI charges 2× the input rate and 1.5× the output rate for the whole request once a prompt passes 272,000 input tokens. Anthropic's Haiku 5.5 steps to $0.50 and $2.50 per million above 100,000 tokens, and xAI bills Grok 4.7 at $4 and $12 once a prompt reaches 200,000. Opus 5.5, Sonnet 5.5, Fable 5.1, Gemini, and Muse charge the same rate across the full window. The table applies these rules automatically and flags the row when one kicks in.

Each row in the price table below links to the vendor page it was read from and shows the date we last checked it. Gemini 4 Argon's $2/$10 is an introductory rate through November 30, 2026; it becomes $4/$20 on December 1, and we will update the table that day.