Models / Gherbal
Language Identification v4

Gherbal

Identify text across 214 languages and 28 writing systems. Gherbal pairs broad language coverage with strong results across Europe, Africa, Asia, and the Americas.

Explore recorded examples
Languages
214
Writing systems
28
8-benchmark mean
83.7%
Model size
200 MB

Gherbal / recorded API replay

One thought. Thirty-one languages.

Choose a passage. Then move across languages and scripts. Every result below came from Gherbal v4.

31 languages · 13 scripts · 3 passages

Developing countries should avoid polluting first and cleaning up later.

Africa / Latin

Yoruba

Hu gba àwọn orílẹ̀èdè tó sẹ̀sẹ̀ ń dàgbà níyànjú láti “kọ ọ̀nà àtijọ́ tí wọ́n má ń ba àyíká jẹ́ saájú kí wọ́n tó tún un se bóbá yá.”
Gherbal identifies
Yorubayor_Latn

Moroccan Arabic / two writing systems

The same language changes script.

Separate recorded API inputs

Latin-script Darija

Salam, labas 3lik? Bghit nsowlek 3la train l Rbat.

→ ary_Latn

Arabic-script Darija

سلام، بغيت نسولك على القطار للرباط.

→ ary_Arab

Passages: FLORES+ devtest, CC BY-SA 4.0, sentence IDs 29, 41, and 48. These are recorded examples, not an accuracy estimate. The benchmark results below use the full evaluation sets.

Read 31 recorded language identifications

Swahili swh_Latn

Hu aliimiza nchi zinazokua “kuepuka njia ya kitambo ya kuchafua kwanza na kusafisha baadaye.’’

Gherbal output: swh_Latn

Yoruba yor_Latn

Hu gba àwọn orílẹ̀èdè tó sẹ̀sẹ̀ ń dàgbà níyànjú láti “kọ ọ̀nà àtijọ́ tí wọ́n má ń ba àyíká jẹ́ saájú kí wọ́n tó tún un se bóbá yá.”

Gherbal output: yor_Latn

Igbo ibo_Latn

Hu gbara obodo ndị na emepe emepe ume "ka ha ghara itinye aka n'usoro omume mgbe ochie bụ ibu ụzọ metọọ were kpochawa ma emechaa."

Gherbal output: ibo_Latn

Zulu zul_Latn

UHu ukhuthaze amazwe athuthukayo "ukugwema indlela yakudala yokugcolisa kuqala bese uhlanza kamuva."

Gherbal output: zul_Latn

Hausa hau_Latn

Hu ya karfafa kasashe masu tasowa “da su guji tsohuwar hanyar gurbatar da farko da kuma tsabtace ta daga baya.”

Gherbal output: hau_Latn

Amharic amh_Ethi

ሑ ታዳጊ ሀገሮች “ውሀን አሁን በክሎ በኋላ የማጥራት መንገድን እንዲተዉ” አበረታተዋል።

Gherbal output: amh_Ethi

Moroccan Arabic ary_Arab

وشجع هو البلدان النامية "باش ما يمشيوش في الطريق القديم اللي كيمشي على مبدأ نلوثوا دابا و ننظفوا من بعد".

Gherbal output: ary_Arab

Standard Moroccan Tamazight zgh_Tfng

ⵉⵙⵙⴱⵖⵙ ⵀⵓ ⵜⵉⵎⵉⵣⴰⵔ ⵉⴱⵓⵖⵍⴰⵏ “ⴰⴷ ⵃⴰⵛⴰⵏⵜ ⴰⵖⴰⵔⴰⵙ ⴰⵣⴰⵢⴽⵓ ⵏ ⵜⵎⵔⴽⵉⵜ ⵣⵡⴰⵔ ⴷ ⵓⵙⵉⵔⴷ ⵙ ⵉⴹⴼⴰⵕ.

Gherbal output: zgh_Tfng

English eng_Latn

Hu encouraged developing countries "to avoid the old path of polluting first and cleaning up later."

Gherbal output: eng_Latn

French fra_Latn

Hu a encouragé les pays en voie de développement « à éviter l'ancienne méthode qui consistait à polluer d'abord et à nettoyer ensuite ».

Gherbal output: fra_Latn

Spanish spa_Latn

Hu incentivó a los estados en vías de desarrollo a «no volver al tradicional camino de contaminar primero y hacer limpieza después».

Gherbal output: spa_Latn

German deu_Latn

Hu ermutigte Entwicklungsländer „den alten Weg von erst verschmutzen und hinterher aufzuräumen, zu vermeiden“.

Gherbal output: deu_Latn

Finnish fin_Latn

Hän kehotti kehitysmaita "välttämään vanhaa tapaa saastuttaa ensin ja siivota jäljet myöhemmin."

Gherbal output: fin_Latn

Turkish tur_Latn

Hu, gelişmekte olan ülkeleri "önce kirletip sonra temizlemeye çalışma şeklindeki eski yöntemden kaçınmaya" teşvik etti.

Gherbal output: tur_Latn

Welsh cym_Latn

Fe wnaeth Hu annog gwledydd datblygol "i osgoi'r hen lwybr o lygru yn gyntaf a glanhau yn nes ymlaen".

Gherbal output: cym_Latn

Irish gle_Latn

Spreag Hu tíortha i mbéal forbartha "chun an seanchosán de bheith ag truailliú ar dtús agus glanadh suas níos déanaí a sheachaint."

Gherbal output: gle_Latn

Russian rus_Cyrl

Ху призывал развивающиеся страны "избегать старой схемы, в которой сначала загрязняют, а потом очищают".

Gherbal output: rus_Cyrl

Ukrainian ukr_Cyrl

Ху закликав країни, що розвиваються "уникати старого шляху, коли спочатку забруднюють, а потім прибирають".

Gherbal output: ukr_Cyrl

Greek ell_Grek

Ο Χου ενθάρρυνε τις αναπυσσόμενες χώρες «να αποφύγουν την παλιά μέθοδο του να ρυπαίνουν πρώτα και να καθαρίζουν αργότερα.»

Gherbal output: ell_Grek

Hindi hin_Deva

हू ने विकासशील देशों को “ पुराने तरीकों से प्रदूषण फैलाने से रोकने और बाद में सफाई करने के लिए प्रोत्साहित किया।”

Gherbal output: hin_Deva

Bengali ben_Beng

"হু উন্নয়নশীল দেশগুলিকে অনুপ্রেরণা দিয়েছিল, ""প্রথমে দূষণের পুরোনো কার্জক্রমগুলিকে পরিহার করা এবং তারপরে পরিষ্কার করা।"""

Gherbal output: ben_Beng

Japanese jpn_Jpan

胡錦濤国家主席は、発展途上国に「まず汚染してから後始末をするというこれまでのやり方を避けるように」と呼びかけました。

Gherbal output: jpn_Jpan

Georgian kat_Geor

ჰუ განვითარებად ქვეყნებს მოუწოდებდა „არ გაიმეორონ ძველი შეცდომები — ჯერ დაბინძურება და შემდეგ დასუფთავება“.

Gherbal output: kat_Geor

Tamil tam_Taml

"வளரும் நாடுகளை ""முதலில் மாசுபடுத்திவிட்டு பின்னர் சுத்தம் செய்வது என்ற பழைய பழக்கத்தை தவிர்க்க"" ஹூ ஊக்குவித்தார். "

Gherbal output: tam_Taml

Vietnamese vie_Latn

Ông khuyến khích các nước đang phát triển nên “tránh đi theo vết xe đổ là gây ô nhiễm trước rồi mới làm sạch sau”.

Gherbal output: vie_Latn

Thai tha_Thai

หูสนับสนุนประเทศกำลังพัฒนาต่าง ๆ ให้ "หลีกเลี่ยงแนวทางเก่าที่เน้นก่อมลพิษเสียก่อนแล้วจึงค่อยแก้ไขในภายหลัง"

Gherbal output: tha_Thai

Hebrew heb_Hebr

"הו עודד מדינות מתפתחות להימנע מדרך הפעולה הישנה שבה קודם מזהמים ואחר כך מנקים."

Gherbal output: heb_Hebr

Urdu urd_Arab

ہو نے ترقی پذیر ممالک کی یہ کہتے ہوئے حوصلہ افزائی کی کہ، "پہلے آلودگی پھیلاؤ اور بعد میں صفائی کرو کے رجحان سے احتراز کریں"

Gherbal output: urd_Arab

Iranian Persian pes_Arab

او کشورهای در حال توسعه را ترغیب کرد «از مسیر قدیمی یعنی ابتدا آلوده کردن محیط زیست و سپس سعی در پاکسازی آن دوری کنند.»

Gherbal output: pes_Arab

Malay zsm_Latn

Hu menggalakkan negara-negara membangun “untuk mengelakkan jalan lama tercemar yang pertama dan membersihkan kemudian.”

Gherbal output: zsm_Latn

Māori mri_Latn

I akiaki a Hu i ngā whenua e whanake ana "kia āraia te huarahi o mua o te parakino i te tuatahi, me te whakapai i muri iho ."

Gherbal output: mri_Latn

One recorded FLORES+ passage per language. An output code can differ from the reference language.

Build with gherbal

Where it fits

See the full stack →

Evaluation / the full picture

Competitive well beyond the Maghreb.

Gherbal leads the eight-benchmark V4-scope mean in the published comparison. General-language sets also show where GlotLID and OpenLID v2 lead.

GHERBAL V483.7%mean accuracy · 200 MB
GLOTLID82.8%mean accuracy · 1,690 MB
OPENLID V282.8%mean accuracy · 1,230 MB

Across language families

Average accuracy by family in the Gherbal v4 report.

Berber / Amazigh97%
Germanic94%
Romance92%
Bantu89%
Indo-Iranian87%
Turkic86%
Atlantic-Congo85%
Austronesian84%
Arabic dialects61%

Eight benchmarks

Accuracy at V4 scope. Each set tests a different distribution.

SetGherbalGlotLIDOpenLID
FLORES devtest92.496.093.9
FLORES dev93.096.893.9
MADAR65.658.865.2
Atlasia69.149.857.4
CommonLID83.390.386.5
WiLI90.694.292.9
Bouquet88.593.992.6
Gherbal-M87.282.779.9

Source: Gherbal v4 Comparative Evaluation Report, March 2026, V4 scope. Scores are percentages. Family results are report summaries; they are not per-language guarantees.

Full coverage

214 Languages, 28 Writing Systems

Every language Gherbal can identify — search by name or code, filter by writing system.

Aa
Aa

Latin

131 languages
AchineseAfrikaansTosk AlbanianMoroccan ArabicAsturianCentral AymaraNorth AzerbaijaniBambaraBalineseBembaBanjarBosnianBugineseCatalanCebuano
ع
ع

Arabic

29 languages
Baharna ArabicAchineseMesopotamian ArabicTa'izzi-Adeni ArabicOmani ArabicTunisian ArabicGulf ArabicLevantine ArabicSudanese ArabicStandard ArabicAlgerian ArabicNajdi ArabicMoroccan ArabicEgyptian ArabicLibyan Arabic
Кк
Кк

Cyrillic

12 languages
BashkirBelarusianBulgarianKazakhHalh MongolianKirghizMacedonianRussianSerbianTatarTajikUkrainian
अ
अ

Devanagari

10 languages
AwadhiBhojpuriHindiChhattisgarhiKashmiriMagahiMaithiliMarathiNepaliSanskrit
অ
অ Bengali
AssameseBengaliManipuri
አ
አ Ethiopic
AmharicTigrinya
ༀ
ༀ Tibetan
TibetanDzongkha
中
中 Chinese
Mandarin ChineseChinese
က
က Myanmar
BurmeseShan
ⵣ
ⵣ Tifinagh
TamasheqStandard Moroccan Tamazight
א
א Hebrew
HebrewEastern Yiddish
Ω
Ω Greek
Modern Greek
અ
અ Gujarati
Gujarati
Ա
Ա Armenian
Armenian
日
日 Japanese
Japanese
ಅ
ಅ Kannada
Kannada
Ⴀ
Ⴀ Georgian
Georgian
អ
អ Khmer
Khmer
ອ
ອ Lao
Lao
അ
അ Malayalam
Malayalam
ଅ
ଅ Odia
Odia
ੴ
ੴ Gurmukhi
Panjabi
ᱚ
ᱚ Ol Chiki
Santali
අ
අ Sinhala
Sinhala
அ
அ Tamil
Tamil
అ
అ Telugu
Telugu
ก
ก Thai
Thai
한
한 Hangul
Korean

Try Gherbal

Send text and get language identification across 214 languages and 28 writing systems.

Open Playground