LLM Token Counter
Count tokens the way OpenAI's tokenizer playground does: colored token spans, a live token/character count, and a Token IDs view. GPT-4o / GPT-4.1 / o-series use the exact o200k_base BPE; GPT-4 / GPT-3.5 use cl100k_base.
Claude, Gemini, Llama, and DeepSeek do not ship a public browser tokenizer. Those counts are labeled ≈ Estimated. Anthropic's official count_tokens API can differ slightly from billed usage. Nothing is sent to an LLM API from this page.
Input
Paste or type your content here…
Results
Paste a prompt to see tokens and IDs.
Frequently Asked Questions
How many tokens is 1000 words?
English is often ~1.3 tokens per word for OpenAI tokenizers, but emoji and CJK text tokenize very differently. Paste your text for an exact GPT-family count.
Why is Claude Estimated?
Anthropic does not publish a client-side tokenizer. Their docs say the count_tokens API is an estimate that can differ slightly from billed usage. This tool approximates Claude with a calibrated OpenAI BPE.
What is a token ID?
Each token has a numeric ID in the model vocabulary. The Token IDs view shows the same array OpenAI's tokenizer playground shows.