NLLB (No Language Left Behind) is a Meta translation model that translates directly between 200 languages, including many with little online text.
In plain words
Translation tools have mostly served a small set of widely used languages, since good systems need lots of paired example sentences. NLLB-200 was built to close that gap. Meta mined new training data for low-resource languages and used a mixture-of-experts model. It was tested on more than 40,000 translation directions and improved quality by 44% on the BLEU score over the previous best. It is released for research, not for production or certified translation.
A quick example
A researcher could give NLLB-200 a single sentence in one of its 200 languages and ask for it in another, then compare the output with a human translation.
Where you'll meet it
- Machine translation research, especially for low-resource languages.
Related
Sources
- PaperNo Language Left Behind: Scaling Human-Centered Machine Translation, NLLB Team (arXiv) · read 29 Sept 2026
- PaperScaling neural machine translation to 200 languages, NLLB Team (Nature) · read 29 Sept 2026
- Docsnllb-200-distilled-600M model card, Meta (Hugging Face) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.