@inproceedings{alastruey-etal-2026-omnilingual,
title = "Omnilingual {MT}: Machine Translation for 1,600 Languages",
author = "Alastruey, Belen and
Bafna, Niyati and
Caciolai, Andrea and
Heffernan, Kevin and
Kozhevnikov, Artyom and
Ropers, Christophe and
S{\'a}nchez, Eduardo and
Saint-James, Charles-{\'E}ric and
Tsiamas, Ioannis and
Cao, Xiang Tony and
Cheng, Chierh and
Chuang, Joe and
Duquenne, Paul-Ambroise and
Duppenthaler, Mark and
Gao, Cynthia and
Huguet Cabot, Pere-Llu{\'i}s and
Janeiro, Jo{\~a}o Maria and
Maillard, Jean and
Mejia-Gonzalez, Gabriel Eduardo and
Schwenk, Holger and
Toledo, Edan and
Turkatenko, Arina and
Ventayol-Boada, Albert and
Arikapudi, Ravi and
Ekberg, Nathanial Paul and
Moritz, Rashel and
Mourachko, Alexandre and
Parimi, Surya and
Williamson, Mary and
Yates, Shireen and
Dale, David and
Costa-juss{\`a}, Marta R.",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1280/",
pages = "29177--29200",
ISBN = "979-8-89176-459-0",
abstract = "We introduce Omnilingual MT, a family of multilingual translation (MT) systems covering more than 1,600 languages{---}the broadest benchmarked coverage of any MT system to date. We make contributions in several key research questions in the field: strategies for the use of monolingual data, curating seed training data, tokenization and suitable architectures, the competitiveness of small specialized MT models as opposed to general purpose large models, and evaluation protocols. Our data strategy combines large-scale multilingual corpora with backtranslation for 2,000+ languages and MeDLEy, a grammatically-diverse seed dataset for 109 low-resource languages. We explore two complementary architectures: OMT-LLAMA, a decoder-only model (1B/3B/8B) built on LLAMA3 with multilingual continual pretraining, and OMT-NLLB, a 3B encoder-decoder model with a novel three-stage training strategy exploiting non-parallel data via autoencoding. For evaluation, we extend the BOUQuET benchmark to 275 languages, develop Met-BOUQuET with human annotations of machine quality in 161 directions, and BLASER 3, a reference-free quality estimation metric. Our specialized 1B{--}8B models match or exceed 70B general-purpose LLMs, demonstrating that specialization is a more reliable path to multilingual translation than scale alone. Omnilingual MT achieves ``passable'' translation quality as judged by human annotators for over 440 languages, doubling it over the previous state of the field at {\ensuremath{\sim}}200 languages, and translates in zero-shot mode into {\ensuremath{\sim}}1,200 languages versus 300{--}400 for the baselines. With Omnilingual MT, we showcase a path encompassing the full pipeline of data and system development towards designing models that are multilingual from the onset."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="alastruey-etal-2026-omnilingual">
<titleInfo>
<title>Omnilingual MT: Machine Translation for 1,600 Languages</title>
</titleInfo>
<name type="personal">
<namePart type="given">Belen</namePart>
<namePart type="family">Alastruey</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Niyati</namePart>
<namePart type="family">Bafna</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andrea</namePart>
<namePart type="family">Caciolai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kevin</namePart>
<namePart type="family">Heffernan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Artyom</namePart>
<namePart type="family">Kozhevnikov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christophe</namePart>
<namePart type="family">Ropers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Eduardo</namePart>
<namePart type="family">Sánchez</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Charles-Éric</namePart>
<namePart type="family">Saint-James</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ioannis</namePart>
<namePart type="family">Tsiamas</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiang</namePart>
<namePart type="given">Tony</namePart>
<namePart type="family">Cao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chierh</namePart>
<namePart type="family">Cheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joe</namePart>
<namePart type="family">Chuang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paul-Ambroise</namePart>
<namePart type="family">Duquenne</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mark</namePart>
<namePart type="family">Duppenthaler</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Cynthia</namePart>
<namePart type="family">Gao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pere-Lluís</namePart>
<namePart type="family">Huguet Cabot</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">João</namePart>
<namePart type="given">Maria</namePart>
<namePart type="family">Janeiro</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jean</namePart>
<namePart type="family">Maillard</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gabriel</namePart>
<namePart type="given">Eduardo</namePart>
<namePart type="family">Mejia-Gonzalez</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Holger</namePart>
<namePart type="family">Schwenk</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Edan</namePart>
<namePart type="family">Toledo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Arina</namePart>
<namePart type="family">Turkatenko</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Albert</namePart>
<namePart type="family">Ventayol-Boada</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ravi</namePart>
<namePart type="family">Arikapudi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nathanial</namePart>
<namePart type="given">Paul</namePart>
<namePart type="family">Ekberg</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rashel</namePart>
<namePart type="family">Moritz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alexandre</namePart>
<namePart type="family">Mourachko</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Surya</namePart>
<namePart type="family">Parimi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mary</namePart>
<namePart type="family">Williamson</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shireen</namePart>
<namePart type="family">Yates</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Dale</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marta</namePart>
<namePart type="given">R</namePart>
<namePart type="family">Costa-jussà</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>We introduce Omnilingual MT, a family of multilingual translation (MT) systems covering more than 1,600 languages—the broadest benchmarked coverage of any MT system to date. We make contributions in several key research questions in the field: strategies for the use of monolingual data, curating seed training data, tokenization and suitable architectures, the competitiveness of small specialized MT models as opposed to general purpose large models, and evaluation protocols. Our data strategy combines large-scale multilingual corpora with backtranslation for 2,000+ languages and MeDLEy, a grammatically-diverse seed dataset for 109 low-resource languages. We explore two complementary architectures: OMT-LLAMA, a decoder-only model (1B/3B/8B) built on LLAMA3 with multilingual continual pretraining, and OMT-NLLB, a 3B encoder-decoder model with a novel three-stage training strategy exploiting non-parallel data via autoencoding. For evaluation, we extend the BOUQuET benchmark to 275 languages, develop Met-BOUQuET with human annotations of machine quality in 161 directions, and BLASER 3, a reference-free quality estimation metric. Our specialized 1B–8B models match or exceed 70B general-purpose LLMs, demonstrating that specialization is a more reliable path to multilingual translation than scale alone. Omnilingual MT achieves “passable” translation quality as judged by human annotators for over 440 languages, doubling it over the previous state of the field at \ensuremath\sim200 languages, and translates in zero-shot mode into \ensuremath\sim1,200 languages versus 300–400 for the baselines. With Omnilingual MT, we showcase a path encompassing the full pipeline of data and system development towards designing models that are multilingual from the onset.</abstract>
<identifier type="citekey">alastruey-etal-2026-omnilingual</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1280/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>29177</start>
<end>29200</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Omnilingual MT: Machine Translation for 1,600 Languages
%A Alastruey, Belen
%A Bafna, Niyati
%A Caciolai, Andrea
%A Heffernan, Kevin
%A Kozhevnikov, Artyom
%A Ropers, Christophe
%A Sánchez, Eduardo
%A Saint-James, Charles-Éric
%A Tsiamas, Ioannis
%A Cao, Xiang Tony
%A Cheng, Chierh
%A Chuang, Joe
%A Duquenne, Paul-Ambroise
%A Duppenthaler, Mark
%A Gao, Cynthia
%A Huguet Cabot, Pere-Lluís
%A Janeiro, João Maria
%A Maillard, Jean
%A Mejia-Gonzalez, Gabriel Eduardo
%A Schwenk, Holger
%A Toledo, Edan
%A Turkatenko, Arina
%A Ventayol-Boada, Albert
%A Arikapudi, Ravi
%A Ekberg, Nathanial Paul
%A Moritz, Rashel
%A Mourachko, Alexandre
%A Parimi, Surya
%A Williamson, Mary
%A Yates, Shireen
%A Dale, David
%A Costa-jussà, Marta R.
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F alastruey-etal-2026-omnilingual
%X We introduce Omnilingual MT, a family of multilingual translation (MT) systems covering more than 1,600 languages—the broadest benchmarked coverage of any MT system to date. We make contributions in several key research questions in the field: strategies for the use of monolingual data, curating seed training data, tokenization and suitable architectures, the competitiveness of small specialized MT models as opposed to general purpose large models, and evaluation protocols. Our data strategy combines large-scale multilingual corpora with backtranslation for 2,000+ languages and MeDLEy, a grammatically-diverse seed dataset for 109 low-resource languages. We explore two complementary architectures: OMT-LLAMA, a decoder-only model (1B/3B/8B) built on LLAMA3 with multilingual continual pretraining, and OMT-NLLB, a 3B encoder-decoder model with a novel three-stage training strategy exploiting non-parallel data via autoencoding. For evaluation, we extend the BOUQuET benchmark to 275 languages, develop Met-BOUQuET with human annotations of machine quality in 161 directions, and BLASER 3, a reference-free quality estimation metric. Our specialized 1B–8B models match or exceed 70B general-purpose LLMs, demonstrating that specialization is a more reliable path to multilingual translation than scale alone. Omnilingual MT achieves “passable” translation quality as judged by human annotators for over 440 languages, doubling it over the previous state of the field at \ensuremath\sim200 languages, and translates in zero-shot mode into \ensuremath\sim1,200 languages versus 300–400 for the baselines. With Omnilingual MT, we showcase a path encompassing the full pipeline of data and system development towards designing models that are multilingual from the onset.
%U https://aclanthology.org/2026.emnlp-main.1280/
%P 29177-29200
Markdown (Informal)
[Omnilingual MT: Machine Translation for 1,600 Languages](https://aclanthology.org/2026.emnlp-main.1280/) (Alastruey et al., EMNLP 2026)
ACL
- Belen Alastruey, Niyati Bafna, Andrea Caciolai, Kevin Heffernan, Artyom Kozhevnikov, Christophe Ropers, Eduardo Sánchez, Charles-Éric Saint-James, Ioannis Tsiamas, Xiang Tony Cao, Chierh Cheng, Joe Chuang, Paul-Ambroise Duquenne, Mark Duppenthaler, Cynthia Gao, Pere-Lluís Huguet Cabot, João Maria Janeiro, Jean Maillard, Gabriel Eduardo Mejia-Gonzalez, Holger Schwenk, Edan Toledo, Arina Turkatenko, Albert Ventayol-Boada, Ravi Arikapudi, Nathanial Paul Ekberg, Rashel Moritz, Alexandre Mourachko, Surya Parimi, Mary Williamson, Shireen Yates, David Dale, and Marta R. Costa-jussà. 2026. Omnilingual MT: Machine Translation for 1,600 Languages. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 29177–29200, Budapest, Hungary. Association for Computational Linguistics.