Free Tool
Token Counter
Count tokens across model families in real time. See how your text splits, compare how each tokenizer counts it, and estimate what a prompt will cost.
Everything runs in your browser. Your text is never uploaded.
Model
Your text
Tokens
37
Characters
180
Words
32
Tokens / word
1.16
Token split
Tokens are the pieces a language model actually reads. A token can be a whole word, part of a word, or a single character. Different model families split the same text differently.
Each colored block is one token for GPT-5.6 / GPT-4o.
The same text across model families
| Model family | Maker | Tokens |
|---|---|---|
| GPT-5.6 / GPT-4o | OpenAI | 37 |
| GPT-4 / GPT-3.5 | OpenAI | 37 |
| Llama 3 | Meta | |
| Qwen 2.5 | Alibaba | |
| Gemma 2 | ||
| Mistral | Mistral | |
| DeepSeek V3 | DeepSeek | |
| Claude | Anthropic | ~43est. |
| Gemini | ~39est. |
Open-model counts use each maker's real tokenizer, loaded on demand. Claude and Gemini are estimated, because neither publishes a public tokenizer; an exact count needs the provider's API. Why tokenizers differ.
Estimated OpenAI input cost
| Model | $ / 1M input | This prompt |
|---|---|---|
| GPT-6 Astra | $5.00 | $0.000185 |
| GPT-5.6 Sol | $2.00 | $0.000074 |
| GPT-5.6 Terra | $1.00 | $0.000037 |
| GPT-5.6 Luna | $0.10 | $0.000004 |
Input cost only, based on 37 tokens under o200k_base. Output tokens are billed separately. Prices as of September 2026; check the official pricing before you rely on a number.
Want to understand what tokens are and why they drive both speed and cost? Read How to Choose a Local LLM.