Free tool
AI Token Counter
Paste text and count tokens for Claude, GPT-6 and Muse models. Exact counts where the tokenizer is verified, labeled estimates everywhere else.
Anthropic
Meta
OpenAI
Only models with verified prices are listed. A model is added after we check the vendor’s price page.
How it works
How this counter works
Exact or estimated
A count is exact only when we have verified the model’s tokenizer and run it locally. For every other model we estimate from characters and say so next to the number — even when another model from the same vendor is exact.
Different models, different counts
The same text can produce a different number of tokens on each model. When you compare costs across models, treat equal token counts as a price scenario, not a promise.
What a count leaves out
An API call also bills your system prompt, conversation history, tool definitions and the output, including reasoning tokens. Use the cost calculator to price the whole call.
Supported models
| Model | Count | Context window |
|---|---|---|
| Claude Opus 5.5 | Estimated | 1,000,000 |
| Claude Sonnet 5.5 | Estimated | 1,000,000 |
| Claude Haiku 5.5 | Estimated | 1,000,000 |
| Muse Wonder | Estimated | 1,048,576 |
| Muse Spark 1.3 | Estimated | 1,048,576 |
| GPT-6.1 Sol | Estimated | 1,050,000 |
| GPT-6 Astra | Estimated | 1,050,000 |
| GPT-6 Luna | Estimated | 1,050,000 |
Models appear here once their price is verified. No local tokenizer is verified yet, so every count is an estimate (characters ÷ 4).