Free tool / cost
How many tokens is that?
Paste text and see the token count and price on GPT, Claude, Gemini, Llama, Mistral, DeepSeek and Grok at once.
No sign-up / runs in your browser
Count once, price everywhere.
GPT-4o and newer
0
Rough estimate
Claude (estimate)
0
Scaled from the real count, about ±15%
Characters
0
0 words
Cheapest to run
—
Paste text to see
Optional. Add the tokens you expect the model to write back to see the full price of one call.
Every model, same text
Prices are US dollars, checked 17 October 2026. Cost is for one call with this text as the prompt.
| Model | Tokens | How counted | Cost of this call | Window used |
|---|---|---|---|---|
| GPT-5.6OpenAI | — | — | — | —of 922K |
| GPT-5.5OpenAI | — | — | — | —of 1.05M |
| GPT-5.4OpenAI | — | — | — | —of 1.05M |
| GPT-5.4 miniOpenAI | — | — | — | —of 272K |
| GPT-5.4 nanoOpenAI | — | — | — | —of 272K |
| GPT-5OpenAI | — | — | — | —of 272K |
| GPT-5 miniOpenAI | — | — | — | —of 272K |
| GPT-5 nanoOpenAI | — | — | — | —of 272K |
| GPT-4.1OpenAI | — | — | — | —of 1.05M |
| GPT-4.1 miniOpenAI | — | — | — | —of 1.05M |
| GPT-4oOpenAI | — | — | — | —of 128K |
| GPT-4o miniOpenAI | — | — | — | —of 128K |
| o3OpenAI | — | — | — | —of 200K |
| o4-miniOpenAI | — | — | — | —of 200K |
| Claude Fable 5.1Anthropic | — | — | — | —of 1M |
| Claude Opus 5.5Anthropic | — | — | — | —of 1M |
| Claude Sonnet 5.5Anthropic | — | — | — | —of 1M |
| Claude Haiku 5.5Anthropic | — | — | — | —of 1M |
| Claude Opus 4.7Anthropic | — | — | — | —of 1M |
| Claude Sonnet 4.6Anthropic | — | — | — | —of 1M |
| Claude Haiku 4.5Anthropic | — | — | — | —of 200K |
| Gemini 3.5 FlashGoogle | — | — | — | —of 1.05M |
| Gemini 3.8 FlashGoogle | — | — | — | —of 1.05M |
| Gemini 3.1 Flash-LiteGoogle | — | — | — | —of 1.05M |
| Gemini 2.5 ProGoogle | — | — | — | —of 1.05M |
| Gemini 2.5 FlashGoogle | — | — | — | —of 1.05M |
| Gemini 2.5 Flash-LiteGoogle | — | — | — | —of 1.05M |
| Llama 4 MaverickMeta | — | — | — | —of 1.05M |
| Llama 4 ScoutMeta | — | — | — | —of 328K |
| Llama 3.3 70BMeta | — | — | — | —of 131K |
| Mistral MediumMistral | — | — | — | —of 262K |
| Mistral LargeMistral | — | — | — | —of 262K |
| Mistral SmallMistral | — | — | — | —of 262K |
| DeepSeek V4 ProDeepSeek | — | — | — | —of 1M |
| DeepSeek V4 FlashDeepSeek | — | — | — | —of 1M |
| DeepSeek ChatDeepSeek | — | — | — | —of 131K |
| Grok 4.5xAI | — | — | — | —of 500K |
| Grok 4.3xAI | — | — | — | —of 1M |
OpenAI (GPT-4o, GPT-4.1, GPT-5 and o-series) counts are exact. Anthropic, Google, Meta, Mistral, DeepSeek and xAI do not publish an offline tokenizer for their current models, so those counts are the real OpenAI count scaled by a typical ratio (for example about 12% more tokens for Claude). Expect roughly ±15% on normal English text, more on code and other languages. Prices are public list prices checked 17 October 2026.
How it works
Three steps, no sign-up.
- 01
Paste your text
A prompt, a document or a whole chat history. It stays in your browser.
- 02
Read the counts
OpenAI counts use the real tokenizer. Other families are scaled from it and clearly labelled as estimates.
- 03
Compare the price
See what one call costs on every model, with an optional reply length, and sort by cheapest.
Questions people ask
Is this the real tokenizer?
For OpenAI models, yes: it runs the open o200k tokenizer in your browser. For every other provider it is an estimate built from that count, and each row says so.
Why do models give different counts for the same text?
Each model family splits text into pieces with its own vocabulary. Code, numbers and non-English text can differ by 20% or more between families, which changes both cost and how much fits in the context window.
Does it include the reply in the price?
Only if you add a reply length. Output tokens cost several times more than input tokens, so a long answer can cost more than the prompt.
Is my text sent to a server?
No. Counting happens in your browser. The only thing downloaded is the tokenizer code.
How do I get exact counts for Claude or Gemini?
Use the provider's own token counting endpoint, or read the usage numbers in each API response. Tracify records those real numbers for every call.
More free tools
- Agent log checker
Paste a log or trace. See where it failed, what was slow and what wasted tokens.
- Model price comparison
Every model's price in one table, plus what your app costs per month on each.
- Prompt cost checker
Paste a prompt. Get a shorter, cheaper version and what it costs on each model.
- Context window checker
Does my document fit in each model's memory?
Now measure the real thing.
Send a first trace and see the true cost, speed and failures of every run.