Token Counter & Prompt Cost Estimator
Paste any prompt, document, or context block and get its exact token count with OpenAI's o200k encoding — then see what sending it costs on every frontier model, from GPT-5.6 and Claude to Gemini, Grok, and DeepSeek.
Counting happens entirely in your browser — your text is never uploaded anywhere.
What this prompt costs per model
Token counts use OpenAI's public o200k_base encoding (GPT-5 / GPT-4o families). Anthropic and Google publish no tokenizer, so rows marked “≈” reuse the same count as an approximation — real counts for those models typically differ by roughly ±10%. Prices are standard per-1M-token API rates, verified July 14, 2026 — full details in the Model Comparison Tracker.
Frequently asked questions
How many tokens is my prompt?
Paste it into the counter above for an exact answer. As a rule of thumb for English text, one token is about 4 characters or about ¾ of a word, so 1,000 words is roughly 1,300–1,500 tokens. Code, non-English text, and unusual formatting tokenize less efficiently and produce more tokens per word.
Do Claude and Gemini count tokens the same way as GPT?
No. Each provider uses its own tokenizer, and OpenAI is the only major provider with a fully public one (the o200k_base encoding used by the GPT-5 and GPT-4o families). Anthropic and Google do not publish theirs, so this tool reuses the o200k count as an approximation for those models — real counts typically differ by roughly ±10%. For billing-exact numbers, both providers return actual token usage in every API response.
How is the cost calculated?
Cost = (input tokens ÷ 1,000,000 × the model's input price) + (expected output tokens ÷ 1,000,000 × the model's output price). Prices are standard per-1M-token API rates verified against official provider pricing pages on July 14, 2026; cached-input, batch, and long-context rates differ.
Is my text uploaded anywhere?
No. The tokenizer runs entirely in your browser via JavaScript — the text you paste never leaves your device, is not sent to any server, and is not stored.
Go deeper
Build with AI? Get the daily rundown.
Original AI briefings for founders, PMs, and builders — free, every morning.