---
title: "Arabic token counter: the real cost of Arabic prompts"
description: "Count tokens in Arabic text across DeepSeek, Qwen, GLM, Gemma and GPT tokenizers and see how many more tokens Arabic needs than English, and what it costs."
url: https://broadrouter.com/tools/arabic-token-counter/
lang: en
updated: 2026-11-15
publisher: BroadRouter (a BroadNet site); API by BroadNet Inference, https://inference.broadnet.ai
---

# Arabic token counter

Updated 2026-11-15

**In short:** Modern Standard Arabic needs 1.31× to 1.77× the tokens of the same sentences in English, depending on the tokenizer (five real tokenizers, our open dataset). An Arabic prompt therefore costs more than its English translation at the same per-token price. Paste Arabic, and its translation if you have one, to see tokens, the ratio and the cost per model.

Free, no sign-up, runs in your browser: nothing you type is sent anywhere.

Samples:

Arabic textيقسّم النموذج اللغوي كل نص يصله إلى وحدات صغيرة تُسمّى الرموز قبل أن يقرأه. قد يكون الرمز كلمة كاملة أو جزءاً من كلمة أو مسافة أو رقماً أو علامة ترقيم. وتُحسب تكلفة الطلب بعدد هذه الرموز، لذلك يدفع المستخدم أكثر كلما احتاج النص إلى رموز أكثر. تحتاج اللغة العربية عادةً إلى رموز أكثر من الإنجليزية لنقل المعنى نفسه، لأن كثيراً من أدوات تقسيم النصوص دُرّبت في الأساس على نصوص إنجليزية. فالكلمة العربية تحمل السوابق واللواحق، مثل أداة التعريف وحروف العطف والضمائر المتصلة، فتتجزأ إلى عدة قطع. وكلما كانت مفردات أداة التقسيم أوسع وأكثر تنوعاً في اللغات، قلّ عدد الرموز التي تحتاجها الجملة العربية. لهذا السبب لا يكفي أن نقدّر التكلفة بقاعدة عامة مأخوذة من النصوص الإنجليزية. الأفضل أن نعدّ الرموز بأداة التقسيم الخاصة بالنموذج الذي سنستخدمه فعلاً، ثم نقارن النتائج بين النماذج المختلفة. قد يكون نموذج أرخص سعراً لكل مليون رمز أغلى في الواقع إذا احتاج إلى ضعف عدد الرموز للنص العربي نفسه. ويُقاس حجم نافذة السياق بالرموز أيضاً. فإذا كان المستند العربي طويلاً، مثل عقد أو تقرير سنوي، فعلينا التأكد من أن النص والتعليمات والإجابة المتوقعة تتسع جميعها داخل حدود النافذة.English translation (optional)Before a language model reads any text, it splits it into small units called tokens. A token may be a whole word, part of a word, a space, a number or a punctuation mark. The cost of a request is calculated by the number of these tokens, so the user pays more whenever the text needs more tokens. Arabic usually needs more tokens than English to carry the same meaning, because many tokenizers were trained mainly on English text. An Arabic word carries prefixes and suffixes, such as the definite article, conjunctions and attached pronouns, so it is broken into several pieces. The wider and more multilingual a tokenizer's vocabulary, the fewer tokens an Arabic sentence needs. For this reason, it is not enough to estimate the cost with a general rule taken from English texts. It is better to count the tokens with the tokenizer of the model we will actually use, and then compare the results between models. A model with a lower price per million tokens may in fact be more expensive if it needs twice as many tokens for the same Arabic text. The size of the context window is also measured in tokens. If the Arabic document is long, such as a contract or an annual report, we must make sure that the text, the instructions and the expected answer all fit within the limits of the window.

| Model | Arabic tokens | English tokens | Arabic tax | Arabic as input | Extra vs English |
| --- | --- | --- | --- | --- | --- |

Arabic tax = Arabic tokens ÷ English tokens for the same meaning. Costs use each model's price per 1M input tokens. ≈ marks an estimate until that tokenizer is loaded.

[Create an account ($5 starting credit)](https://inference.broadnet.ai/?utm_source=broadrouter&utm_medium=referral&utm_campaign=tools-arabic-token-counter&utm_content=tool) ·

[Report a wrong price or number](mailto:info@broadnet.ai?subject=Wrong%20price%20or%20number%20on%20Arabic%20token%20counter&body=Page%3A%20https%3A%2F%2Fbroadrouter.com%2Ftools%2Farabic-token-counter%2F%0A%0AWhat%20looks%20wrong%20(model%2C%20figure%20shown%2C%20correct%20figure%20and%20a%20source%20link%20if%20you%20have%20one)%3A%0A) (email to info@broadnet.ai, with this page's link and your inputs filled in).

## How it works

1.  Paste Arabic text into the right-to-left box, or pick a sample. Optionally paste its English translation in the second box.
2.  The table counts tokens with each model's tokenizer. Load the real tokenizers to replace the estimates marked ≈.
3.  Read the Arabic-to-English ratio (the "Arabic tax") and the extra cost of sending the Arabic version as input.
4.  Copy the link to share the comparison, or download CSV.

## The Arabic tax, measured

We counted 6,076 Arabic sentences, in Modern Standard Arabic and three dialects, and their English translations with each real tokenizer: 2,009 FLORES-200 sentences each for MSA, Egyptian and Levantine, and 49 Tatoeba sentences for Gulf Arabic, which FLORES-200 does not cover (so the Gulf column is less precise). Arabic tax = Arabic tokens ÷ English tokens for the same sentences.

| Tokenizer | MSA tax | Gulf tax | Egyptian tax | Levantine tax | MSA tokens per word | Input price per 1M |
| --- | --- | --- | --- | --- | --- | --- |
| [DeepSeek V4.1 Flash](https://broadrouter.com/models/deepseek-v4.1-flash/) | **1.62×** | 1.44× | 1.63× | 1.58× | 2.22 | $0.175 |
| [GLM 5.3](https://broadrouter.com/models/glm-5.3/) | **1.77×** | 1.57× | 1.75× | 1.70× | 2.44 | $0.78 |
| [Qwen3.8 27B](https://broadrouter.com/models/qwen3.8-27b/) | **1.31×** | 1.21× | 1.35× | 1.29× | 1.83 | $0.175 |
| [Gemma 4 VL 26B](https://broadrouter.com/models/gemma-4-vl-26b/) | **1.47×** | 1.22× | 1.44× | 1.41× | 2.03 | $0.10 |
| GPT (o200k) | **1.38×** | 1.32× | 1.41× | 1.35× | 1.89 | — |

Measured 2026-11-15. Parallel text: FLORES-200 dev and devtest (2,009 aligned sentences per language, CC BY-SA 4.0, tarball pinned by SHA-256) for Modern Standard, Egyptian, North Levantine, Najdi and Mesopotamian Arabic, Hindi, Urdu, Chinese, French and Spanish against English, each sentence counted on its own; Tatoeba sentence pairs (CC BY 2.0 FR) for Gulf Arabic and Arabizi. Ratio = pooled tokens ÷ pooled tokens of the English side, with a 95% interval from a seeded bootstrap over sentences (10,000 resamples). Other texts: Wikipedia lead sections (CC BY-SA 4.0), Project Gutenberg (public domain) and Wikidata JSON (CC0), stored with source URL, licence and revision. Tokens: the official tokenizer file at a pinned Hugging Face commit, encoded with @huggingface/tokenizers with no special tokens; o200k\_base via gpt-tokenizer; Kimi's tiktoken rank file with a tiktoken-compatible counter. Gated Meta and Google files were read from ungated mirrors proven byte-identical to the official file. Every new count matched the Rust tokenizers library or Python tiktoken. Words: word-like segments from Intl.Segmenter. Characters: Unicode code points. Every text, source and count is in the [tokenizer efficiency dataset](https://broadrouter.com/data/tokenizer-efficiency/) (CSV and JSON).

On Modern Standard Arabic the tax ranges from 1.31× (Qwen3.8 27B) to 1.77× (GLM 5.3). English counts are close across tokenizers, so the difference comes from how much Arabic each vocabulary learned. Dialects land near MSA, but dialect spellings that a vocabulary never saw split into more pieces.

## The formula

```
Arabic tax          = Arabic tokens ÷ English tokens (same meaning)
Arabic input cost   = Arabic tokens × input price per 1M ÷ 1,000,000
extra vs English    = (Arabic tokens − English tokens) × input price per 1M ÷ 1,000,000
```

## Worked example

The news-style MSA sample in the tool is 316 tokens with the Qwen3.8 27B tokenizer and its English translation is 269. On [Qwen3.8 27B](https://broadrouter.com/models/qwen3.8-27b/) at $0.175 per 1M input tokens, sending the Arabic costs $0.000055 against $0.000047 for English: 17% more for the same meaning. Across a million such requests that is $8.22 extra.

## How to pay less Arabic tax

-   Pick the model by cost per Arabic request, not by price per token. The table above does that multiplication.
-   Keep system prompts and few-shot examples in English when the model follows them well, and send only user content in Arabic.
-   Strip tashkeel and tatweel (ـ) from text the model only needs to read; normalise Arabic-Indic digits if you send many numbers.
-   Cache repeated instructions: cached input is billed at a fraction of the price. See [cached input tokens](https://broadrouter.com/blog/cached-input-tokens-explained/).

Prices synced 2026-11-14. Google's Gemma 4 26B A4B tokenizer. BroadNet's build is based on this model; we assume it keeps the base tokenizer.

## Questions

### How many tokens is an Arabic word?

Between 1.8 and 2.4 tokens per word of Modern Standard Arabic in our open dataset, depending on the tokenizer, against 1.3 to 1.3 for English prose. Dialects are similar. Paste your own text above for the exact figure.

### Why does Arabic use more tokens than English?

Tokenizers learn their vocabulary from training text, and most training text is English. An Arabic word also carries attached pieces, such as the article ال, the conjunction و and pronoun suffixes, so one written word often maps to several tokens. Bigger multilingual vocabularies learn more whole Arabic words and need fewer tokens.

### Which BroadNet model is cheapest for Arabic text?

It depends on both the token count and the price. On Modern Standard Arabic sentences Qwen3.8 27B needs the fewest tokens relative to English (1.31×). Multiply each model's Arabic token count by its price, as the table above does, to compare the real cost. [Choosing an LLM for Arabic](https://broadrouter.com/blog/choosing-an-llm-for-arabic/) covers quality too.

### Do Arabic diacritics (tashkeel) add tokens?

Yes, a lot. Each harakah is a separate Unicode character, and vocabularies rarely learn vowelled words. In our test the sentence ذهب الولد إلى المدرسة is 5 to 9 tokens across these tokenizers, and the fully vowelled ذَهَبَ الوَلَدُ إِلَى المَدْرَسَةِ is 14 to 27. Strip tashkeel from text the model only needs to read.

### Are Arabic-Indic digits counted differently from Western digits?

Yes. ١٢٣٤٥ and 12345 are different characters. In our test 12345 is 2 tokens with DeepSeek, GLM and GPT and 5 with Qwen and Gemma, which split every digit, while ١٢٣٤٥ is 5 to 10 tokens. Normalise digits if you send many numbers.

### Is my text sent anywhere?

No. Tokenization runs in your browser. The only download is the public tokenizer file from Hugging Face, once; your text stays on your device.

## Go deeper

-   [Open data: tokenizer efficiency across languages (CSV, JSON)](https://broadrouter.com/data/tokenizer-efficiency/)
-   [Open data: real API cost per language, by model and provider](https://broadrouter.com/data/real-cost-by-language/)
-   [Choosing an LLM for Arabic](https://broadrouter.com/blog/choosing-an-llm-for-arabic/)
-   [Qwen3.8 27B: tested in Arabic by BroadNet](https://broadrouter.com/models/qwen3.8-27b/)
-   [LLM token counter for every model](https://broadrouter.com/tools/token-counter/)
-   [Read Arabic invoices with a vision model](https://broadrouter.com/guides/read-arabic-invoices-with-vision/)
-   [What is a token?](https://broadrouter.com/glossary/token/)

## More free tools

-   [**LLM API cost calculator**Price a workload on every model at once, with cache hits, reasoning and a flat-plan check.](https://broadrouter.com/tools/llm-api-cost-calculator/)
-   [**Speech-to-text cost calculator**Hours, calls or voice notes to a monthly transcription bill, against your current provider.](https://broadrouter.com/tools/speech-to-text-cost-calculator/)
-   [**LLM token counter**Exact token counts with each model's real tokenizer, with a token view and cost.](https://broadrouter.com/tools/token-counter/)
-   [**Tokens to words converter**Tokens ↔ words ↔ characters ↔ pages, by language and content type.](https://broadrouter.com/tools/tokens-to-words/)
-   [**Context window calculator**Will it fit? Window used per model, room left for output, cost to send.](https://broadrouter.com/tools/context-window-calculator/)
-   [**Flat-rate vs per-token LLM cost calculator**Break-even between an unlimited plan and paying per token, per agent.](https://broadrouter.com/tools/flat-rate-vs-per-token-calculator/)
-   [**Prompt caching savings calculator**Bill with and without caching, by prefix share and hit rate.](https://broadrouter.com/tools/prompt-caching-calculator/)
-   [**Image token calculator**Tokens and cost per image under each model's documented rule. Drop an image; nothing is uploaded.](https://broadrouter.com/tools/image-token-calculator/)
-   [**Self-hosting vs API cost calculator**GPU rental against your current API and BroadNet per-token and flat-plan prices: cost per 1M tokens and break-even volume.](https://broadrouter.com/tools/self-hosting-vs-api-calculator/)
-   [**Embedding cost calculator**Index, update and query costs for a corpus at OpenAI, Google, Voyage, Mistral and Jina list prices.](https://broadrouter.com/tools/embedding-cost-calculator/)
-   [**LLM rate limit and concurrency calculator**Which limit binds first: RPM, TPM or concurrency, with Little's law.](https://broadrouter.com/tools/llm-rate-limit-calculator/)
-   [**JSON schema generator for LLM structured output**Sample JSON to strict schema, Pydantic, Zod and a ready request.](https://broadrouter.com/tools/json-schema-generator/)
-   [**OpenAI-compatible API request builder**One request, copied as curl, Python, Node or the OpenAI SDKs.](https://broadrouter.com/tools/api-request-builder/)
-   [**Audio file splitter for speech-to-text limits**Chunk count, overlap and copy-ready ffmpeg commands to fit a transcription file-size limit.](https://broadrouter.com/tools/audio-splitter-ffmpeg/)
-   [**Agent run cost estimator**Context growth per step, total tokens and cost of one agent run, with a DeepSeek session comparison.](https://broadrouter.com/tools/agent-run-cost-estimator/)
-   [**OpenAI to BroadNet migration helper**Paste OpenAI SDK code; get base URL, key and model edits, a diff and the checks that remain.](https://broadrouter.com/tools/openai-to-broadnet-migration/)
-   [**Arabic text normalizer**Tashkeel, tatweel, alef, ya, ta marbuta, digits and invisible characters, with a matching Python function.](https://broadrouter.com/tools/arabic-text-normalizer/)
-   [**RAG chunking calculator**Chunks, embedding tokens, input per query, context check and monthly LLM cost for a RAG pipeline.](https://broadrouter.com/tools/rag-chunking-calculator/)
-   [**max\_tokens calculator for reasoning models**Answer length + reasoning budget + margin = a max\_tokens that does not cut the answer off.](https://broadrouter.com/tools/reasoning-token-budget/)
-   [**LLM batch job planner**Run time, binding limit, cost per model and the flat sessions needed to finish a big job by a deadline.](https://broadrouter.com/tools/llm-batch-job-planner/)
-   [**LLM JSON output validator**Parse errors by line and column, safe repairs, cut-off detection and JSON Schema checks for model output.](https://broadrouter.com/tools/llm-json-output-validator/)
-   [**Prompt token diff**Two prompt versions side by side: changed lines, exact token delta and the monthly cost difference.](https://broadrouter.com/tools/prompt-token-diff/)

Try it on your real workload. New accounts start with $5 of credit for Gemma and transcription. DeepSeek, GLM and Qwen3.8 need a top-up.

[Get started](https://inference.broadnet.ai/account/?utm_source=broadrouter&utm_medium=referral&utm_campaign=tools-arabic-token-counter&utm_content=page-cta) ·

**Sources**

-   [BroadNet Inference machine-readable prices (pricing.json)](https://inference.broadnet.ai/pricing.json?utm_source=broadrouter&utm_medium=referral&utm_campaign=tools-arabic-token-counter&utm_content=sources)
-   [DeepSeek V4.1 Flash tokenizer (deepseek-ai/DeepSeek-V4.1-Flash)](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash)
-   [GLM 5.3 tokenizer (zai-org/GLM-5.3)](https://huggingface.co/zai-org/GLM-5.3)
-   [Qwen3.8 27B tokenizer (Qwen/Qwen3.8-27B)](https://huggingface.co/Qwen/Qwen3.8-27B)
-   [Gemma 4 tokenizer (Google base model) (google/gemma-4-26B-A4B-it)](https://huggingface.co/google/gemma-4-26B-A4B-it)
-   [gpt-tokenizer (o200k\_base)](https://github.com/niieani/gpt-tokenizer)
