T-LM enhances GPT-4's top-notch performance in 200 languages

Translated Language Model (T-LM) enables businesses creating content in languages other than English to fully leverage the text generation capabilities of GPT-4. The service addresses the performance and cost disparity associated with using GPT-4 in languages other than English, enabling companies to create and restructure content in 200 languages.

Use Cases

Assisting global content
creation teams

T-LM assists content creators in generating and restructuring content in 200 languages.

Enhancing multilingual
customer support

T-LM enables chatbots and support systems to operate smoothly in the user's language.

Facilitating user-generated content
creation on global platforms

T-LM helps users generate content on multilingual platforms in their native language

Every other use case of GPT-4 with prompts in languages other than English.

Why T-LM

Until now, GPT's impressive performance has been a privilege of the English-speaking world. Companies operating in languages other than English have often found their performance lagging behind that of GPT models from several years ago, with some languages trailing by as much as three years. For these companies, the performance gap in understanding, generating, and restructuring content in languages other than English was an ongoing challenge that often prevented them from taking full advantage of generative AI.

Additionally, using GPT-4 in languages other than English can cost up to 15 times more because the pricing model is based on text segmentation (tokenization) that is optimized for English.

How T-LM works

The disparity in GPT-4’s performance between English and other languages arises from the predominance of English-centric sources – such as the Common Crawl dataset and Wikipedia – in training data, leading to inferior outcomes in languages other than English. T-LM addresses this disparity by translating the initial prompt from the source language to English then back to the user's language using a specialized model of ModernMT.

Clients can optionally use their existing ModernMT keys to employ adaptive models within GPT-4. At no additional cost, they can also submit reference materials like website content, knowledge bases, or brand guidelines to adapt both the MT and language model components.

This approach also lowers the cost of using GPT-4 by reducing the number of token required to process the prompts.

Start using T-LM

Give your company the full potential of GPT-4 in 200 languages.

Request API access

The following list comprises the languages for which we recommend using our solution.


Acehnese - ace

Afrikaans - af

Akan - ak

Albanian - sq

Amharic - am

Arabic - ar

Armenian - hy

Assamese - as

Asturian - ast

Awadhi - awa

Ayacucho Quechua - quy

Aymara, Central - ayr

Azerbaijani - az

Azerbaijani, Northern - azi

Azerbaijani, Southern - azb


Balinese - ban

Bambara - bm

Banjar - bjn

Bashkir - ba

Belarusian - be

Bemba - bem

Bengali - bn

Bhojpuri - bho

Bosnian - bs

Buginese - bug

Bulgarian - bg


Catalan - ca

Cebuano - ceb

Chhattisgarhi - hne

Chinese (Simplified) - zh-CN

Chinese (Traditional) - zh-TW

Chokwe - cjk

Crimean Tatar - crh

Croatian - hr

Czech - cs


Danish - da

Dari - prs

Dimli - diq

Dinka, Southwestern - dik

Dutch - nl

Dyula - dyu

Dzongkha - dz


English - en

Esperanto - eo

Estonian - et

Ewe - ee


Faroese - fo

Fijian - fj

Finnish - fi

Fon - fon

French - fr

Friulian - fur


Galician - gl

Ganda - lg

Georgian - ka

German - de

Greek - el

Guarani - gn

Gujarati - gu


Haitian - ht

Halh Mongolian - khk

Hausa - ha

Hebrew - he

Hindi - bjn

Hungarian - hu


Icelandic - is

Igbo - ig

Iloko - ilo

Indonesian - id

Irish - ga

Italian - it


Japanese - ja

Javanese - jv

Jingpho - kac


Kabiyè - kbp

Kabuverdianu - kea

Kabyle - kab

Kamba - kam

Kannada - kn

Kanuri, Central (Latin script) - knc

Kashmiri (Arabic script) - kas

Kashmiri (Devanagari script) - ks

Kazakh - kk

Khmer - km

Kikuyu - ki

Kimbundu - kmb

Kinyarwanda - rw

Kongo - kg

Korean - ko

Kurdish, Central - ckb

Kurdish, Northern - kmr

Kyrgyz - ky


Lao - lo

Latgalian - ltg

Latin - la

Latvian - lv

Ligurian - lij

Limburgish - li

Lingala - ln

Lithuanian - lt

Lombard - lmo

Luba-Kasai - lua

Luo - luo

Luxembourgish - lb


Macedonian - mk

Magahi - mag

Maithili - mai

Malagasy - mg

Malay - ms

Malayalam - ml

Maltese - mt

Manipuri - mni

Maori - mi

Marathi - mr

Minangkabau - min

Mizo - lus

Marathi - mr

Minangkabau - min

Mizo - lus

Mongolian (Traditional) - mn

Mossi - mos

Myanmar (Burmese) - my


Nepali - ne

Nigerian Fulfulde - fuv

Norwegian Bokmål - nb

Norwegian Nynorsk - nn

Nuer - nus

Nyanja - ny


Occitan - oc

Oriya - or

Oromo, West Central - gaz


Pangasinan - pag

Papiamento - pap

Pashto, Southern - pbt

Pastho - ps

Persian, Western - pes

Plateau Malagasy - plt

Polish - pl

Portuguese (Brazilian) - pt-BR

Portuguese (European) - pt-PT

Punjabi - pa


Romanian - ro

Rundi - rn

Russian - ru


Samoan - sm

Sango - sg

Sanskrit - sa

Santali - sat

Sardinian - cs

Scots Gaelic - gd

Serbian (Cyrillic) - sr-Cyrl

Serbian (Latin) - sr-Latn

Shan - shn

Shona - sn

Sicilian - scn

Silesian - szl

Sindhi - sd

Sinhala (Sinhalese) - si

Slovak - sk

Slovenian - sl

Somali - so

Northern Sotho - nso

Southern Sotho - st

Spanish - es-ES

Spanish (Latin America) - es-419

Standard Latvian - lvs

Standard Malay - zsm

Sundanese - su

Swahili - sw

Swati - ss

Swedish - sv


Tagalog - tl

Tajik - tg

Tamasheq - taq

Tamazight, Central Atlas - tzm

Tamil - ta

Tatar - tt

Telugu - te

Thai - th

Tibetan - bo

Tigrinya - ti

Tok Pisin - tpi

Tosk Albanian - als

Tsonga - ts

Tswana - tn

Tumbuka - tum

Turkish - tr

Turkmen - tk

Twi - tw


Ukrainian - uk

Umbundu - umb

Urdu - ur

Uyghur - ug

Uzbek, Northern - uzn


Venetian - vec

Vietnamese - vi


Waray (Philippines) - war

Welsh - cy

Wolof - wo


Xhosa - xh


x - x

Yoruba - yo


Zulu - zu

• T-LM has the same limitation as GPT-4 in English.
• Using English as a pivot language, T-LM is unable to answer questions that require a specific understanding of a country's culture.
• The performance of T-LM on the MMLU benchmark may not be representative of performance in other domains or tasks.

Get in touch.

We are here to answer your questions, and help you get what you want.

Contact us