Token & Cost Visualizer
Paste any text to estimate its tokens, see how it splits, price it at your model’s rate, and check it against common context windows — entirely in your browser.
Your text
Type or paste below — everything updates live.
How it splits
Each coloured piece is one estimated token.
Estimate, not the exact tokenizer. Real token counts depend on the model’s own BPE and vary ±10–20%. Coloured pieces approximate the split. For billing-critical numbers, verify with the model’s tokenizer. Context-window sizes are dated June 2026 — confirm with the provider.
Good to know
Is the token count exact?
No — it’s a transparent estimate. A true count needs the specific model’s tokenizer (BPE), which we deliberately don’t call out to a third party. The estimate is typically within ~10–20%; for billing-critical numbers, check with the model’s own tokenizer.
Why approximate instead of the real tokenizer?
Real tokenizers ship large model-specific data and many tools send your text to a server to run them. Keeping it in-browser and dependency-free means your text never leaves the page — the whole point of these tools.
Roughly how many tokens per word?
For English, ~1.3 tokens per word (≈ 4 characters per token). Code, other languages and lots of punctuation push the ratio up.
Want this guaranteed in your own infrastructure?
Get a copy of these results by email and see RAGSuite running on a setup like yours — citation-backed, self-hosted, EU-ready.
Thanks — we’ll be in touch within one business day (DE/EN).
Tokens add up — predictability doesn’t have to.
RAGSuite runs on a flat self-hosted instance and supports local models via Ollama, so heavy retrieval context doesn’t turn into a runaway per-token bill.