A token counter estimates how many tokens a prompt or document uses under a model’s tokenizer so you can plan context limits and API cost before you call a model. OmniKit counts in the browser with selectable encodings and can send totals into the LLM Cost Calculator. Counts are for pasted text; chat wrappers and tools may add overhead.
Paste a prompt and count tokens in this tab. OpenAI GPT-4o and GPT-5-class text uses tiktoken o200k_base; GPT-4 / GPT-3.5 uses cl100k_base. Claude and Gemini are cl100k approximations and are labeled as such. Nothing is uploaded. Open the LLM Cost Calculator when you need monthly dollars, or Prompt Token Diff when you are comparing two prompts.
Required fields on the left
Matches tiktoken o200k_base used by GPT-4o and GPT-5-class chat models.
Appears after you run
Paste text to count tokens
OpenAI encodings use tiktoken in the browser. Claude and Gemini are labeled approximations.
Chat wrappers, tools, and system prefixes add overhead the API bills and this paste does not include. Compare two prompts on Prompt Token Diff.
A token counter estimates how many tokens a prompt or document uses under a model’s tokenizer so you can plan context limits and API cost before you call a model. OmniKit counts in the browser with selectable encodings and can send totals into the LLM Cost Calculator. Counts are for pasted text; chat wrappers and tools may add overhead.
OpenAI families use tiktoken o200k_base or cl100k_base in this tab (checked 12 September 2026). Claude and Gemini rows are cl100k stand-ins and are labeled approximations.
On 12 September 2026 OmniKit’s Token Counter uses js-tiktoken o200k_base for GPT-4o / GPT-5-class pastes and cl100k_base for GPT-4 / GPT-3.5. Claude and Gemini rows reuse cl100k and are labeled approximations. Counts run in the browser; chat templates can still change the API invoice.
Paste text, pick an encoding family, and read tokens, words, and characters. OpenAI rows use js-tiktoken in the browser. Claude and Gemini reuse cl100k and stay marked approx. Estimate monthly $ opens the cost calculator with a share of the paste (truncated if the URL would be too long).
This page does not call OpenAI, Anthropic, or Google. It does not spend AI credits. Message formatting, tools, and cached prefixes can still change the invoice.
Paste the prompt or document.
Pick an encoding family. Read the approximation note for Claude and Gemini.
Copy the count, or open Estimate monthly $ to prefill the cost calculator.
Providers tokenize the chat template, tool schemas, and sometimes images. This page only encodes the string you paste. Treat the number as a planning floor, then confirm on the vendor usage dashboard.
No. Token Counter is one paste and one encoding. Prompt Token Diff compares two prompts on tokens and monthly dollars across models. Use the counter first, then the diff when you are choosing a rewrite.
Use OpenAI GPT-4o / GPT-5 family (o200k_base). GPT-4 and GPT-3.5 use cl100k_base. Those two rows are tiktoken-exact for the pasted string. Claude and Gemini are not.
OpenAI tiktoken encodings o200k_base and cl100k_base; OmniKit check date 12 September 2026.
A token counter estimates how many tokens a prompt or document uses under a model’s tokenizer so you can plan context limits and API cost before you call a model. OmniKit counts in the browser with selectable encodings and can send totals into the LLM Cost Calculator. Counts are for pasted text; chat wrappers and tools may add overhead.
For o200k_base and cl100k_base, yes — those rows use js-tiktoken, the same encoding names as OpenAI’s tokenizer. Official platform.openai.com/tokenizer remains the vendor page if you need their UI.
The API tokenizes the full message list, tools, and sometimes images. This page only encodes the paste. Use it as a planning floor.
No. Tokenization runs in the tab. Estimate monthly $ only puts a truncated paste into a share URL on the cost calculator if it fits.
Keep measuring in the same cluster — or jump to the next decision.
Start here when the question is “how long is this prompt in tokens?” Price the same paste on the LLM Cost Calculator (catalog v2026-09-11). Compare two system prompts on Prompt Token Diff. Read encodings on this page — last checked 12 September 2026.
