---
title: "What is Modern Standard Arabic (MSA) vs Arabic dialects?"
description: "Modern Standard Arabic is the shared written and formal Arabic of media and education. How MSA differs from dialects and why it matters for LLMs and speech-to-text."
url: https://broadrouter.com/glossary/msa/
lang: en
updated: 2026-11-16
publisher: BroadRouter (a BroadNet site); API by BroadNet Inference, https://inference.broadnet.ai
---

# Modern Standard Arabic (MSA)

Updated 2026-11-16

Modern Standard Arabic (MSA) is the standardized form of Arabic used in writing, news, education and formal speech across the Arab world. It is nobody's native everyday language: people speak regional dialects such as Egyptian or Gulf Arabic and switch to MSA in formal settings.

Also called: Modern Standard Arabic, Fusha, Fus'ha, Standard Arabic, al-fusha

When people say an AI model “speaks Arabic”, they usually mean it handles Modern Standard Arabic. Dialects are a separate test.

## How it works

Arabic is a family of varieties. The MADAR paper (Bouamor et al., 2018) puts it plainly: MSA is “the shared language of culture, media, and education from Morocco to Oman”, yet it “is not the native language of any speaker of Arabic”. Dialects are grouped regionally as Egyptian, North African, Levantine, Gulf and Yemeni, and sub-regionally as Tunisian, Lebanese, Kuwaiti, Qatari and so on. MADAR itself covers 25 city dialects alongside MSA.

In practice the two mix. In unscripted settings where MSA would be expected, such as TV talk shows, speakers switch back and forth between MSA and their dialect. That is [code-switching](https://broadrouter.com/glossary/code-switching/), and the 2014 code-switching shared task found MSA versus dialect the hardest pair to tell apart at the word level, harder than Spanish-English or Nepali-English.

|  | MSA | Dialects |
| --- | --- | --- |
| Where | news, books, school, contracts, formal speech | home, street, phone calls, voice notes, chat |
| Spelling | standard | no fixed spelling; often typed as [Arabizi](https://broadrouter.com/glossary/arabizi/) |
| Training data | large written corpora | much less, unevenly spread by region |

## Why it matters

-   **LLM quality.** A model can write good MSA and still miss Gulf or Maghrebi slang. Test with the dialect your users speak.
-   **Speech-to-text accuracy.** Many test sets are read or broadcast speech, while your audio may be dialect voice notes. Ask for [word error rate](https://broadrouter.com/glossary/word-error-rate/) by variety, not one Arabic number, and test on your own recordings.
-   **Token cost.** In our [tokenizer dataset](https://broadrouter.com/data/tokenizer-efficiency/), on [Qwen3.8 27B](https://broadrouter.com/models/qwen3.8-27b/) Modern Standard Arabic needed 1.31× the tokens of the same sentences in English (2,009 FLORES-200 sentences) and Gulf Arabic 1.21× (a small set of 49 Tatoeba sentences, so a rough figure). Use the [Arabic token counter](https://broadrouter.com/tools/arabic-token-counter/) on your own text.

## On BroadNet Inference

BroadNet reports testing Qwen3.8 27B for fluent Arabic writing, with Gulf and Egyptian dialects understood. [BroadNet Transcribe Arabic](https://broadrouter.com/models/cohere-transcribe-arabic/) is listed for Arabic including dialects. For both MSA and dialect audio, send `language=ar`; BroadNet’s docs warn that without it Arabic can come back in Latin letters. The [dialect speech recognition article](https://broadrouter.com/blog/arabic-dialect-speech-recognition/) explains how to test on your own recordings.

## Questions

### Which language code should I send for dialect audio?

Send ar. BroadNet's docs say to send language=ar for Arabic, and BroadNet Transcribe Arabic is listed for Arabic including dialects, so there is no separate code per dialect.

### Do Arabic speakers understand MSA?

Educated speakers read and understand it, since it is the language of school, books and news. In everyday conversation they speak their dialect.

### Which language code is MSA?

ISO 639-3 lists Standard Arabic as arb, part of the Arabic macrolanguage ara. Speech APIs usually take the two-letter code ar for all Arabic.

## See also

-   [Code-switching](https://broadrouter.com/glossary/code-switching/) (term)
-   [Arabizi](https://broadrouter.com/glossary/arabizi/) (term)
-   [Speech-to-text](https://broadrouter.com/glossary/speech-to-text/) (term)
-   [Word error rate](https://broadrouter.com/glossary/word-error-rate/) (term)
-   [Tokenizer](https://broadrouter.com/glossary/tokenizer/) (term)
-   [Dialectal Arabic](https://broadrouter.com/glossary/dialectal-arabic/) (term)
-   [Arabic diacritics (tashkeel)](https://broadrouter.com/glossary/arabic-diacritics/) (term)
-   [Arabic speech-to-text API, step by step](https://broadrouter.com/guides/arabic-speech-to-text-api/) (guide)
-   [Gulf, Egyptian and Levantine Arabic: how LLMs handle dialects](https://broadrouter.com/blog/arabic-llm-gulf-egyptian-dialects/) (article)
-   [Arabic dialect speech recognition: Gulf, Egyptian, Levantine, Maghrebi](https://broadrouter.com/blog/arabic-dialect-speech-recognition/) (article)
-   [Choosing an LLM API for Arabic: what to test](https://broadrouter.com/blog/choosing-an-llm-for-arabic/) (article)

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=glossary-msa&utm_content=page-cta) ·

**Sources**

-   [Bouamor et al., The MADAR Arabic Dialect Corpus and Lexicon (LREC 2018, ACL Anthology) (checked 2026-11-16)](https://aclanthology.org/L18-1535/)
-   [Solorio et al., Overview for the First Shared Task on Language Identification in Code-Switched Data (2014, ACL Anthology) (checked 2026-11-16)](https://aclanthology.org/W14-3907/)
-   [SIL International, ISO 639-3 code arb (Standard Arabic) (checked 2026-11-16)](https://iso639-3.sil.org/code/arb)
-   [BroadNet Inference API reference, transcription (language parameter) (checked 2026-11-16)](https://inference.broadnet.ai/docs?utm_source=broadrouter&utm_medium=referral&utm_campaign=glossary-msa&utm_content=sources)
