Token counter
Paste a text. The number appears immediately as an estimate per model family. Your text does not leave the browser.
The counter runs in the browser and needs JavaScript. Without JavaScript, use the rule of thumb from the table.
| Model family | EN | DE · FR · IT · ES · PT |
|---|---|---|
| OpenAI GPT-4o, GPT-4.1, GPT-5, gpt-oss | 4.2 | 4 |
| Anthropic Claude | 3.5 | 2.9 |
| Google Gemini, Gemma | 4 | 3.5 |
| Meta Llama | 3.9 | 3.3 |
| Mistral | 3.8 | 3.3 |
| Qwen, DeepSeek, Kimi | 3.8 | 3 |
How we estimate: characters divided by an average per family and language. For technical text, code and rare languages the real number deviates more. Example: 1,000 characters of German text give about 250 tokens.