Context Budget

Paste a prompt, code, or docs → estimate tokens and see if it fits the window

In-browser estimate, nothing uploaded. Tokens are estimated; tokenizers vary by model (~±15%).

0 chars0 words0 CJK

How it estimates

  1. Counts CJK vs other characters separately
  2. Applies rough ratios: ~4 chars/token for English, ~1.5 tokens/char for CJK
  3. Maps onto each model window with a ±15% band
  4. No model call, no network. Local and instant.