Tokenizer & Payload inspector
No API key · Your input stays in this browser
Loading tokenizer…
Tokenizer
Paste a prompt, code, or multilingual text. Counts update as you type.
01Choose an encoding and an example
Try an example
02Enter your text
—UTF-8 · Up to 100,000 UTF-16 code units
03Inspect the token sequence
Each color marks a token. Select one to see its ID and decoded fragment.
Spaces → · Newlines → ↵
A token may contain only part of a UTF-8 character and display � on its own. Decoding the complete sequence restores the text. The view shows up to 2,000 tokens; counts and exports include all tokens.
04Compare or export
Ready · Text tokens only; API message overhead is excluded.
Where does your context go?
Add each input section, reserve output space, then inspect the largest contributor.
01Break down your input
02Set your budget
03Review usage and remaining space
Sections are encoded separately and summed. Real requests add separators, message templates, tools, and possibly multimodal content. Confirm limits for the target model and compare with API usage.
From request structure to media.
Paste JSON → inspect field sizes → decode and preview embedded media locally.
01Paste request JSON
Up to 12 MiB UTF-8 · External URLs are never fetched automatically
02Review request summary
Paste a complete JSON request or load the working media example.
03Explore fields and preview media
Field sizes measure the UTF-8 size of each serialized JSON value. Parent and child sizes overlap. Base64 expands binary data by about one third. Byte size is not a measure of model media tokens.