Skip to content
SPECIFICTION
Get listed

NLLB-200

ai.meta.com

WebSelf-hosted
Visit ai.meta.com
Spec score 7.6 #103 of 158 in Translation apps

Spec sheet

Purpose
NLLB-200 is an open-source AI project with models for direct translation between 200 languages, including low-resource languages.Read 13 Oct 2026 · ai.meta.com
Intended users
Meta presents the released models and research tools as resources for researchers and communities building or improving translation tools.Read 13 Oct 2026 · ai.meta.com
License
The 3.3B model card lists the license as CC-BY-NC-4.0.Read 12 Oct 2026 · huggingface.co
Headquarters
Menlo Park, California, United StatesRead 28 Sep 2026 · ai.meta.com
Founded
2004Read 28 Sep 2026 · ai.meta.com
Certified translations
The model card says NLLB-200 translations cannot be used as certified translations.Read 12 Oct 2026 · huggingface.co
Code license
The repository says NLLB code and fairseq are MIT-licensed.Read 13 Oct 2026 · github.com
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.Read 12 Oct 2026 · huggingface.co
Document translation
NoRead 22 Sep 2026 · ai.meta.com
Evaluation
Meta says it created the FLORES-200 dataset to evaluate translation quality across the supported languages.Read 13 Oct 2026 · ai.meta.com
Input length
The model card says training inputs did not exceed 512 tokens and longer sequences may have lower translation quality.Read 12 Oct 2026 · huggingface.co
Integration
Dense models are available through Hugging Face Hub and supported by the Transformers library.Read 13 Oct 2026 · github.com
Language coverage
Meta says the model includes 150 low-resource languages.Read 12 Oct 2026 · ai.meta.com
Library support
The 3.3B model card provides usage instructions for the Transformers library and links to Google Colab and Kaggle notebooks.Read 12 Oct 2026 · huggingface.co
Model license
The repository states that the NLLB models are licensed under CC-BY-NC 4.0.Read 13 Oct 2026 · github.com
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.Read 13 Oct 2026 · github.com
Open-source release
Meta says it open-sourced NLLB-200 models, FLORES-200, model training code, and code for recreating the training dataset.Read 12 Oct 2026 · ai.meta.com
Reported quality
Meta reports that NLLB-200 exceeded the previous state of the art by an average of 44% in its evaluations.Read 13 Oct 2026 · ai.meta.com
Repository status
The fairseq repository was archived on March 20, 2026, and is read-only.Read 13 Oct 2026 · github.com
Research resources
The project releases model training and inference code, evaluation benchmarks, language identification models, LASER3 encoders, and data mining code.Read 13 Oct 2026 · github.com
Research use
The model card describes NLLB-200 as primarily intended for machine translation research, especially for low-resource languages.Read 12 Oct 2026 · huggingface.co
Safety evaluation
Meta describes evaluating translations against human translations and filtering profanity and other offensive content using toxicity lists for supported languages.Read 13 Oct 2026 · ai.meta.com
Translation directions
The project says its models support more than 200 × 200 translation directions.Read 13 Oct 2026 · github.com
Translation quality
Meta reports an average 44% BLEU improvement over the previous state of the art.Read 12 Oct 2026 · ai.meta.com
Use case
Meta says Wikipedia editors use NLLB technology through Wikimedia’s Content Translation tool to translate articles in low-resource languages.Read 13 Oct 2026 · ai.meta.com
Use limits
The model card says it is not released for production deployment and is not intended for domain-specific or document translation.Read 12 Oct 2026 · huggingface.co
Wikipedia integration
Wikipedia editors use NLLB-200 technology through the Wikimedia Foundation’s Content Translation Tool for more than 20 low-resource languages.Read 12 Oct 2026 · ai.meta.com

Where Specifiction places it

Alternatives to NLLB-200All alternatives →