Token counter: how many tokens is my text?
Paste any text to estimate its tokens and what it costs to send to current AI models, with a reply of the length you choose. Prices checked 13 Oct 2026.
What it costs to send
Cheapest first
| Model | Once | 1,000 times |
|---|---|---|
| Ministral 3 3BMistral AI · $0.10 / $0.10 | $0.0001 | $0.057 |
| Ministral 3 8BMistral AI · $0.15 / $0.15 | $0.0001 | $0.086 |
| Ministral 3 14BMistral AI · $0.20 / $0.20 | $0.0001 | $0.11 |
| Qwen3.8 FlashAlibaba · $0.15 / $0.47 | $0.0002 | $0.25 |
| Claude Haiku 5.5Anthropic · $0.10 / $0.50 | $0.0003 | $0.26 |
| GPT-6 LunaOpenAI · $0.10 / $0.50 | $0.0003 | $0.26 |
| GLM-5.3 FlashZ.ai · $0.15 / $0.50 | $0.0003 | $0.26 |
| Mistral Small 4Mistral AI · $0.15 / $0.60 | $0.0003 | $0.31 |
| DeepSeek V4.1 FlashDeepSeek · $0.30 / $1.20 | $0.0006 | $0.62 |
| MiniMax M3MiniMax · $0.30 / $1.20 | $0.0006 | $0.62 |
| GLM-5.3 FlashXZ.ai · $0.37 / $1.25 | $0.0007 | $0.65 |
| Command A+Cohere · $0.30 / $1.50 | $0.0008 | $0.77 |
| Mistral Large 3Mistral AI · $0.50 / $1.50 | $0.0008 | $0.79 |
| Qwen3.7 PlusAlibaba · $0.40 / $1.60 | $0.0008 | $0.83 |
| Gemini 3.5 Flash-LiteGoogle · $0.30 / $2.50 | $0.0013 | $1.27 |
| Nova 2 LiteAmazon · $0.30 / $2.50 | $0.0013 | $1.27 |
| Gemini 3.8 FlashGoogle · $0.75 / $3.75 | $0.0019 | $1.93 |
| DeepSeek V4 ProDeepSeek · $1.32 / $3.96 | $0.0021 | $2.08 |
| Muse Spark 1.3Meta · $1.25 / $4.25 | $0.0022 | $2.22 |
| GLM-5.3Z.ai · $1.40 / $4.40 | $0.0023 | $2.30 |
| Grok 4.7xAI · $2.00 / $6.00 | $0.0031 | $3.15 |
| Qwen3.8 MaxAlibaba · $2.00 / $6.00 | $0.0031 | $3.15 |
| Mistral Medium 3.5Mistral AI · $1.50 / $7.50 | $0.0039 | $3.86 |
| Claude Sonnet 5.5Anthropic · $2.00 / $10.00 | $0.0051 | $5.15 |
| GPT-6.1 SolOpenAI · $2.00 / $10.00 | $0.0051 | $5.15 |
| Gemini 3.1 ProGoogle · $2.00 / $12.00 | $0.0061 | $6.15 |
| Kimi K3Moonshot AI · $3.00 / $15.00 | $0.0077 | $7.72 |
| Claude Opus 5.5Anthropic · $4.00 / $20.00 | $0.010 | $10.30 |
| Claude Fable 5.1Anthropic · $10.00 / $50.00 | $0.026 | $25.74 |
| GPT-6 AstraOpenAI · $10.00 / $50.00 | $0.026 | $25.74 |
List prices for current general-purpose models, no caching or batch discount; long-prompt rates apply where a model charges more above a length. Token counts are estimates: Claude, GPT and Gemini tokenizers each split text differently. Price a monthly workload
What is a token?
AI models don’t read letters or whole words. They read tokens: chunks of text from a fixed vocabulary, often a whole short word with its leading space (“ the”), a piece of a longer word (“token” + “izer”), a number of up to three digits, or a punctuation mark. Providers bill per million tokens, separately for the tokens you send (input) and the tokens the model writes (output).
Rules of thumb for English
| Text | About |
|---|---|
| 1 token | 4 characters, or ¾ of a word |
| 100 tokens | 75 words |
| A page of text (500 words) | 670 tokens |
| A 300-page book (90,000 words) | 120,000 tokens |
| 1 million tokens | 750,000 words, or 8 novels |
These are averages for plain English prose. Code, tables, numbers and most other languages use more tokens per word.
From tokens to a bill
Cost is tokens times price. On a model at $2 / $10 per million input / output tokens, a 2,000-token prompt with a 500-token reply costs about $0.01; a million such requests cost about $9,000. Repeated input can be much cheaper with prompt caching, and batch jobs are often half price. The token cost calculator handles monthly volumes, caching and batch.
Questions
How many tokens is a word?
In English, about 1.3 tokens per word on average, or roughly 4 characters per token: 100 tokens is about 75 words. Common words are a single token; long, rare and non-English words take more. Code, numbers and languages such as Chinese, Japanese and Korean use more tokens for the same amount of text.
Is this count exact?
No, it’s an estimate. Every provider uses its own tokenizer, so the same text comes to a different number of tokens on Claude, GPT and Gemini models. For an exact count, use the provider’s own tool: OpenAI publishes its tokenizer (tiktoken), and Anthropic and Google offer token-counting endpoints in their APIs.
Does my text leave my browser?
No. The count and the prices are worked out in your browser; nothing you paste is sent anywhere.
Why do output tokens cost more?
Generating text takes more computing than reading it, so most providers charge several times more for output than for input. A long answer can cost more than a long prompt.
What else adds tokens?
In a chat, every earlier message is sent again with each new one, so a conversation’s input grows with every turn. System prompts, tool definitions, images and files count as input too. Reasoning models also bill their hidden reasoning as output tokens.