LLM Token Counter
NewEstimate token counts for GPT, Claude, and Llama-style models as you type.
LLM token counter
Count tokens for GPT, Claude, and Llama-style models
Paste text to see an estimated token count per model family, updated as you type. Estimates use character/word heuristics, not the exact vendor tokenizer.
Frequently Asked Questions
Is this the exact token count OpenAI or Anthropic would bill me for?
No. It blends a character-based estimate (~3.3-4 chars/token depending on model family) with a word-count estimate to approximate real tokenizer behavior. For exact billing counts, use the vendor's own tokenizer (e.g. tiktoken for OpenAI).
Why do different models show different counts for the same text?
Each model family uses a different chars-per-token ratio in the estimate — GPT models assume about 4 characters per token, Claude about 3.5, and Llama/Mistral about 3.3, reflecting typical differences in vocabulary density.
Does it update as I type, or do I need to click a button?
It updates automatically about 150ms after you stop typing. There is no Count button.
Does any of my text get sent anywhere?
No. Everything runs in your browser using string length and word-split math — no network request is made.
Will this help me stay under a model's context window?
It gives a reasonable planning estimate against the context windows listed on the page (e.g. 128k for GPT-4o), but leave headroom since it is an approximation, not an exact count.
Related Tools
AI Cost Calculator
NewEstimate LLM API costs from input/output token counts across popular models.
Prompt Template Builder
NewBuild reusable AI prompt templates with {{variable}} placeholders and instant preview.
JSON to TOON
NewConvert JSON to TOON, a compact token-oriented format that reduces token usage in LLM prompts.