← Back to Tools
Token counter · cross-tokenizer
Paste text to see its token count and cost under each tokenizer. Counts for the OpenAI family are computed exactly in the browser; the rest are marked as estimates with a margin. The spread is widest for Chinese and other CJK text, where o200k needs noticeably fewer tokens than cl100k for the same characters.
0 characters
Tokens across tokenizers
| Tokenizer | Tokens | Chars / token | Accuracy |
|---|---|---|---|
GPT-4o / GPT-5 (o200k) | 0 | — | Exact |
GPT-4 / GPT-3.5 (cl100k) | 0 | — | Exact |
Claude | 0 | — | Estimate ±20% |
Gemini | 0 | — | Estimate ±20% |
Llama 3 | 0 | — | Estimate ±25% |
Exact counts use gpt-tokenizer (o200k_base / cl100k_base). Estimates anchor on the exact o200k count and scale by CJK ratio, so treat them as ballpark figures rather than precise ones. Claude and Gemini have no public client-side tokenizer.
Cost for this text, by model
Cost = tokens ÷ 1M × the model's list price. “As input” uses the input price, “as output” the output price. Token counts for non-OpenAI models are estimates.