Token Explorer
Compare token counts across LLMs
Type or paste text to see how leading model families tokenize it. Results update live, and exact token delineations are rendered where available. See How it works for implementation details around client side tokenizers, provider APIs, and supported models.
15 characters37 UTF-8 bytes
Drag this handle to reorder, or focus it and use the up and down arrow keys. Home and End move a model to the start or end of the list. Manual reordering changes the Order by selection to manual.
GPT-5
OpenAI7 tokens
- 오늘149830
- 도5827
- ␠좋은70135
- ␠하루141945
- ␠보내129466
- 세요37436
- !0
o200k_base · Local · [email protected]
Meta Llama 4
Meta7 tokens
- 오늘115380
- 도2762
- ␠좋은48191
- ␠하루83197
- ␠보내105218
- 세요28350
- !13
llama4/tokenizer.model · Local · meta-llama/llama-models@0e0b8c51
Grok 4.7
xAI10 tokens
- 오늘61081
- 도2061
- ␠884
- 좋은66221
- ␠884
- 하루66473
- ␠884
- 보내60967
- 세요34787
- !885
grok-4.7 · provider API
Claude Opus 5.5
AnthropicToken breakdown unavailable — Anthropic's API returns counts only.
claude-opus-5-5 · provider API
The displayed number is an approximate raw-content count from differential measurements (observed range 18–18). Anthropic reports 25 tokens for the complete structured input. Actual token IDs and boundaries are unavailable.