FREEDOM251

Freedom251 › Translation apps

NLLB-200

by NLLB-200

Free plan Self-hostedWeb

Visit NLLB-200 Alternatives

NLLB-200's own site

What NLLB-200 is

NLLB-200 is an open-source machine translation model that translates directly among 200 languages, including 150 described as low-resource. The project supports more than 200 × 200 translation directions, and Meta reports an average 44% BLEU improvement over the previous state of the art in its evaluations. Releases include model files, the FLORES-200 benchmark, training and inference code, dataset-recreation code, language-identification models, LASER3 encoders, and data-mining code. Dense models are available through Hugging Face Hub and the Transformers library, with Google Colab and Kaggle notebooks. Variants include a 54.5B-parameter MoE model, dense 3.3B and 1.3B models, and distilled 1.3B and 600M models. The project is free and available for self-hosting or web use. It is intended mainly for translation research, especially low-resource languages. The 3.3B model uses CC-BY-NC-4.0, while code uses MIT licensing. It is not released for production, domain-specific, or document translation.

Who it suits

NLLB-200 suits researchers and communities building translation tools, particularly for low-resource languages. It is not intended for production deployment or certified translation work.

What it does well

  • Translates directly among 200 languages
  • Includes 150 low-resource languages
  • Open-source models and research code
  • Multiple model sizes and variants
  • Works with Transformers and Hugging Face

What to know first

  • Not released for production deployment
  • Not intended for document translation
  • Longer inputs may reduce quality
  • Translations cannot serve as certified translations

Verdict

NLLB-200 offers broad multilingual coverage and extensive research resources at no charge. Its research-only positioning, licensing, input limits, and possible mistranslations require careful use.

Facts from NLLB-200's own pages

Certified translations
The model card says NLLB-200 translations cannot be used as certified translations. · huggingface.co, 12 Oct 2026
Code license
The repository says NLLB code and fairseq are MIT-licensed. · github.com, 13 Oct 2026
Data and accuracy caveats
The model card warns that personally identifiable information may not be entirely eliminated from training data and that mistranslations may remain. · huggingface.co, 12 Oct 2026
Document translation
No · ai.meta.com, 22 Sept 2026
Evaluation
Meta says it created the FLORES-200 dataset to evaluate translation quality across the supported languages. · ai.meta.com, 13 Oct 2026
Founded
2004 · ai.meta.com, 28 Sept 2026
Headquarters
Menlo Park, California, United States · ai.meta.com, 28 Sept 2026
Input length
The model card says training inputs did not exceed 512 tokens and longer sequences may have lower translation quality. · huggingface.co, 12 Oct 2026
Integration
Dense models are available through Hugging Face Hub and supported by the Transformers library. · github.com, 13 Oct 2026
Intended users
Meta presents the released models and research tools as resources for researchers and communities building or improving translation tools. · ai.meta.com, 13 Oct 2026
Language coverage
Meta says the model includes 150 low-resource languages. · ai.meta.com, 12 Oct 2026
Library support
The 3.3B model card provides usage instructions for the Transformers library and links to Google Colab and Kaggle notebooks. · huggingface.co, 12 Oct 2026
License
The 3.3B model card lists the license as CC-BY-NC-4.0. · huggingface.co, 12 Oct 2026
Model license
The repository states that the NLLB models are licensed under CC-BY-NC 4.0. · github.com, 13 Oct 2026
Model variants
The repository lists NLLB-200 MoE at 54.5B parameters and dense versions at 3.3B and 1.3B, plus distilled versions at 1.3B and 600M. · github.com, 13 Oct 2026
Open-source release
Meta says it open-sourced NLLB-200 models, FLORES-200, model training code, and code for recreating the training dataset. · ai.meta.com, 12 Oct 2026
Purpose
NLLB-200 is an open-source AI project with models for direct translation between 200 languages, including low-resource languages. · ai.meta.com, 13 Oct 2026
Reported quality
Meta reports that NLLB-200 exceeded the previous state of the art by an average of 44% in its evaluations. · ai.meta.com, 13 Oct 2026
Repository status
The fairseq repository was archived on March 20, 2026, and is read-only. · github.com, 13 Oct 2026
Research resources
The project releases model training and inference code, evaluation benchmarks, language identification models, LASER3 encoders, and data mining code. · github.com, 13 Oct 2026
Research use
The model card describes NLLB-200 as primarily intended for machine translation research, especially for low-resource languages. · huggingface.co, 12 Oct 2026
Safety evaluation
Meta describes evaluating translations against human translations and filtering profanity and other offensive content using toxicity lists for supported languages. · ai.meta.com, 13 Oct 2026
Translation directions
The project says its models support more than 200 × 200 translation directions. · github.com, 13 Oct 2026
Translation quality
Meta reports an average 44% BLEU improvement over the previous state of the art. · ai.meta.com, 12 Oct 2026

Where NLLB-200 stands on Freedom251

Look at these too