See how Gemma 3 splits plain text into tokens, inspect token IDs, and keep the text in your browser.
Your text stays in this browser. Nothing you paste is uploaded.
Plain text is encoded locally with the Gemma 3 tokenizer. Chat templates and model-specific processing are not added.
This page runs the pinned Gemma 3 SentencePiece tokenizer locally in a Web Worker and returns token pieces and IDs.
Only plain text is encoded. BOS, EOS, chat templates, roles, system instructions, tools, and media tokens are not added.
For request-volume and output-token calculations, see the monthly LLM API cost guide. Using OpenAI? Check the current OpenAI API token prices.
To compare the same English, Chinese, JSON, and code samples, open the reproducible tokenizer comparison.
Gemma 3 tokenizer sourceHigh confidenceChecked 2026-07-19
StackLens pins the Gemma 3 SentencePiece tokenizer asset and runs it locally without model weights, chat templates, or special-token formatting.
Yes. The tokenizer asset is loaded into a browser Worker, and the text you enter is not uploaded by StackLens.
Yes. The result exposes the token pieces and IDs produced by the pinned Gemma 3 tokenizer.
No. This page encodes plain text without BOS, EOS, roles, system instructions, or a chat template.
Not necessarily. Model-specific request formatting and chat templates can add tokens beyond the plain text shown here.