Claude Tokenizer
The Claude Tokenizer counts the exact number of tokens your text uses for Anthropic's Claude models, using Anthropic's official token-counting API. Paste your text, pick a model, and get an accurate count instead of an estimate.
Why not use a GPT / tiktoken tokenizer for Claude?
Claude does not use OpenAI's tokenizer, so counting Claude tokens with tiktoken gives the wrong number. Older "Claude tokenizer" websites use the legacy Claude 1/2 tokenizer, which is also wrong for modern Claude 3 and 4 models. The only accurate count for a current model is the one Anthropic's own API produces — which is exactly what this tool returns.
How to count Claude tokens
- Paste or type your text.
- Choose the Claude model you are targeting (Opus, Sonnet, or Haiku).
- Select Count tokens to get the exact count for that model.
Claude models and tokenizer generations
| Model | Tokenizer |
|---|---|
| Claude Opus 4.8 · Sonnet 4.5 · Haiku 4.5 | Current (Opus 4.7+) |
| Claude Opus 4.6 | Older (pre-4.7) |
Anthropic changed the tokenizer starting with Claude Opus 4.7, so the same text can produce a slightly different count on Opus 4.6 versus Opus 4.8. Count against the model you actually call.
Example
The quick brown fox jumps over the lazy dog. counts to about 12 tokens on Claude Opus 4.8. Longer or non-English text uses more tokens per character — use the live characters-per-token readout to gauge your own text.
Limitations
- A free sign-in is required; counting tokens costs no credits.
- No per-token colored breakdown: Anthropic's tokenizer isn't public, so this tool returns the exact count only. For a visual token view, use the ChatGPT Tokenizer (OpenAI's tokenizer is open).
- Counts are model-specific — switch models to see how the number changes.
