Token counter.
Paste any text and get an estimated token count plus the live cost on every tracked model, cheapest first, with a context-fit check for each. No email gate, no signup — that is the whole pitch.
How the estimate works: a blended heuristic — the average of characters ÷ 4 and words × 1.33, the two standard rules of thumb for English prose. It is an estimate, not any one model’s exact tokenizer; code and non-English text can differ by 10–20%. Costs use live list prices from the database.
Words ↔ tokens converter
Rule of thumb for English prose: tokens ≈ words × 1.33.
By model family
- Claude token counter — 7 models
- Command token counter — 2 models
- DeepSeek token counter — 2 models
- Gemini token counter — 4 models
- Gemma token counter — 2 models
- GPT token counter — 7 models
- Grok token counter — 3 models
- Inkling token counter — 2 models
- Jamba token counter — 2 models
- Kimi token counter — 3 models
- Llama token counter — 2 models
- MAI token counter — 2 models
- Mistral token counter — 4 models
- Muse token counter — 3 models
- Nemotron token counter — 3 models
- Nova token counter — 2 models
- Qwen token counter — 3 models
FAQ
What is a token?
A token is the unit AI models read and write — roughly three-quarters of an English word. Models price by the token and limit context by the token, so counting tokens is counting both cost and capacity.
How many tokens is 1,000 words?
About 1,330 tokens for typical English prose (words × 1.33). Dense or technical text runs higher; the counter above estimates from your actual text.
How much does 1 million tokens cost?
It depends entirely on the model: in our database the blended rate (3:1 input:output) runs from $0.20 per 1M on GPT-6 Luna to $20.00 per 1M on Claude Fable 5.1 — which is why the same prompt can cost 100× more on one model than another. Full table: pricing index.
How do I reduce token costs?
Trim system prompts, cache repeated context where the maker offers cached-input pricing, summarise long documents before the expensive call, cap output length, and route easy work to a cheaper model — the cost calculator shows the saving per workload.