Model Description

MiLMMT-46-12B-Pretrain is a language model developed through continual pretraining of Gemma3-12B using a mix of 143 billion tokens from both monolingual and parallel data across 46 different languages. Please find more details in our paper: Scaling Model and Data for Multilingual Machine Translation with Open Large Language Models.

Supported Languages: Arabic, Azerbaijani, Bulgarian, Bengali, Catalan, Czech, Danish, German, Greek, English, Spanish, Persian, Finnish, French, Hebrew, Hindi, Croatian, Hungarian, Indonesian, Italian, Japanese, Kazakh, Khmer, Korean, Lao, Malay, Burmese, Norwegian, Dutch, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Swedish, Tamil, Thai, Tagalog, Turkish, Urdu, Uzbek, Vietnamese, Cantonese, Chinese (Simplified), Chinese (Traditional).
GitHub: Please find more details in our GitHub repository.
Developed by: Xiaomi Inc.

Note that MiLMMT-46-12B-Pretrain is NOT a translation model.

Training Data

We collect monolingual data from DCAD-2000. For parallel data, we collect all Chinese-centric and English-centric parallel datasets from the OPUS collection up to August 2025 and conduct a series of filtering processes, such as language identification, semantic duplication filtering, quality filtering, and more.

Citation

@misc{shang2026scalingmodeldatamultilingual,
      title={Scaling Model and Data for Multilingual Machine Translation with Open Large Language Models}, 
      author={Yuzhe Shang and Pengzhi Gao and Wei Liu and Jian Luan and Jinsong Su},
      year={2026},
      eprint={2602.11961},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2602.11961}, 
}