How many tokens is 1,000 words?

For ordinary English prose, about 1,300 tokens with OpenAI's current tokenizers. Code, JSON, numbers and non-Latin scripts use more tokens per word, so paste your actual text to get its real count.

Which tokenizer does GPT-4o use?

GPT-4o, GPT-4.1, GPT-5 and the o-series reasoning models use o200k_base. GPT-4, GPT-3.5 Turbo and the text-embedding-3 models use the older cl100k_base. This page counts both exactly.

Why does Amharic use so many tokens?

Tokenizers are built from the text they were trained on, which is mostly English and other Latin-script languages. Ge'ez characters appear rarely, so most are stored as two or three byte-level tokens instead of being merged into whole words. The page shows the characters-per-token ratio when it detects Ge'ez script.

Is my text sent anywhere?

No. The tokenizers load into your browser once and count locally, so prompts and documents never leave the page.

LLM token counter

Paste a prompt or a document to count its tokens for OpenAI's models, with an estimate for Claude. Each token is highlighted so you can see how the model splits your text, including why Amharic and code cost more than plain English.

Free
No signup
Runs in your browser
o200k + cl100k exact

Your text or prompt

Tokens in your text

Loading tokenizer…

GPT-5, GPT-4.1, GPT-4o, o-series

o200k_base · exact

…

GPT-4, GPT-3.5 Turbo, text-embedding-3

cl100k_base · exact

…

Claude

tokenizer not public · rough estimate

…

166

Characters

Words

–

Chars / token

Counted in your browser with the same tokenizers OpenAI's models use. A chat API call adds a few tokens per message for roles and formatting.

EthioDeploy

Calling an LLM API from your app? Keep the key on a server.

An API key in front-end code is readable by every visitor. Put the call behind a small backend instead: EthioDeploy runs a Node.js or Python API from GitHub with the key stored as an encrypted environment variable, on a free plan with no card.

Host a backend free

How to use it

01
Paste your text
A system prompt, a document for retrieval, or a whole chat. Counts update as you type.
02
Read the right row
GPT-4o and newer OpenAI models use o200k_base; GPT-4, GPT-3.5 and the text-embedding-3 models use cl100k_base. Both counts are exact.
03
Check the split
Switch the highlight between tokenizers to see where words break. Hover a chunk to see how many tokens it took.

Things that change the count

Language and script: English averages around four characters per token. Scripts with fewer training examples, such as Ge'ez for Amharic and Tigrinya, can take one or more tokens per character, so the same message costs several times more.
Whitespace and formatting: Indentation, repeated spaces and pretty-printed JSON all cost tokens. Minify JSON you send as context, and drop blank lines from logs before pasting them into a prompt.
Chat overhead: A chat API call wraps each message with role markers, so a conversation costs a few tokens per message beyond its text. Tool definitions and images are counted separately.
Claude: Anthropic doesn't publish the tokenizer for current Claude models, so the Claude row is a rough range based on the o200k count. For an exact number, call Anthropic's token counting endpoint (/v1/messages/count_tokens) with your prompt.

FAQ

Questions

Last updated November 1, 2026

A token is the unit a language model reads and writes: a common word, part of a longer word, a punctuation mark or a few bytes of a rarer character. API pricing and context limits are both measured in tokens, not characters or words.

All tools

Deploy.env validator & converterFind duplicate keys, broken quotes and leaked public secrets, then convert to JSON, YAML, Docker or Compose.BotsDiscord permissions calculatorPick bot permissions, get the integer and the invite link, or decode a permissions number.

Ship it from GitHub.

EthioDeploy builds and runs your app on every push, with a database and HTTPS. Free plan, no card.

Deploy free

LLM token counter

Free
No signup
Runs in your browser
o200k + cl100k exact

Your text or prompt

Tokens in your text

Loading tokenizer…

GPT-5, GPT-4.1, GPT-4o, o-series

o200k_base · exact

…

GPT-4, GPT-3.5 Turbo, text-embedding-3

cl100k_base · exact

…

Claude

tokenizer not public · rough estimate

…

166

Characters

Words

–

Chars / token

Counted in your browser with the same tokenizers OpenAI's models use. A chat API call adds a few tokens per message for roles and formatting.

EthioDeploy

Calling an LLM API from your app? Keep the key on a server.

Host a backend free

How to use it

01
Paste your text
A system prompt, a document for retrieval, or a whole chat. Counts update as you type.
02
Read the right row
GPT-4o and newer OpenAI models use o200k_base; GPT-4, GPT-3.5 and the text-embedding-3 models use cl100k_base. Both counts are exact.
03
Check the split
Switch the highlight between tokenizers to see where words break. Hover a chunk to see how many tokens it took.

Things that change the count

Language and script: English averages around four characters per token. Scripts with fewer training examples, such as Ge'ez for Amharic and Tigrinya, can take one or more tokens per character, so the same message costs several times more.
Whitespace and formatting: Indentation, repeated spaces and pretty-printed JSON all cost tokens. Minify JSON you send as context, and drop blank lines from logs before pasting them into a prompt.
Chat overhead: A chat API call wraps each message with role markers, so a conversation costs a few tokens per message beyond its text. Tool definitions and images are counted separately.
Claude: Anthropic doesn't publish the tokenizer for current Claude models, so the Claude row is a rough range based on the o200k count. For an exact number, call Anthropic's token counting endpoint (/v1/messages/count_tokens) with your prompt.

FAQ

Questions

Last updated November 1, 2026

All tools

Ship it from GitHub.

EthioDeploy builds and runs your app on every push, with a database and HTTPS. Free plan, no card.

Deploy free

LLM token counter

Calling an LLM API from your app? Keep the key on a server.

How to use it

Paste your text

Read the right row

Check the split

Things that change the count

Questions

What is a token?+

How many tokens is 1,000 words?+

Which tokenizer does GPT-4o use?+

Why does Amharic use so many tokens?+

Is my text sent anywhere?+

More free tools

Ship it from GitHub.

LLM token counter

Calling an LLM API from your app? Keep the key on a server.

How to use it

Paste your text

Read the right row

Check the split

Things that change the count

Questions

What is a token?+

How many tokens is 1,000 words?+

Which tokenizer does GPT-4o use?+

Why does Amharic use so many tokens?+

Is my text sent anywhere?+

More free tools

Ship it from GitHub.