MiLMMT-46-4B-v1.0-EXL3-4.0bpw

Model Description

MiLMMT-46-4B-v1.0 is an LLM-based translation model. Its training builds on Gemma3-4B through four stages: (1) continual pretraining on 143 billion tokens of monolingual and parallel data across 46 languages, yielding MiLMMT-46-4B-Pretrain; (2) supervised finetuning, yielding MiLMMT-46-4B-v0.1; and (3) reinforcement learning; and (4) model merging, yielding this model, MiLMMT-46-4B-v1.0.

• 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.

Model Performance

Translation Prompt

Translate this from <source language name> to <target language name>:
<source language name>: <source language sentence>
<target language name>:

Please use the language name specified above in the translation prompt.

Run the model

Using on vLLM:

from vllm import LLM, SamplingParams


model_id = "xiaomi-research/MiLMMT-46-4B-v1.0"

model = LLM(model=model_id)
sampling_params = SamplingParams(top_k=1, temperature=0, max_tokens=2048)

text = "Translate this from Chinese (Simplified) to English:\nChinese (Simplified): 我爱机器翻译\nEnglish:"

outputs = model.generate(text, sampling_params)
print(outputs[0].outputs[0].text)

Using on Transformers:

from transformers import AutoModelForCausalLM, AutoTokenizer


model_id = "xiaomi-research/MiLMMT-46-4B-v1.0"

model = AutoModelForCausalLM.from_pretrained(model_id)
tokenizer = AutoTokenizer.from_pretrained(model_id)

text = "Translate this from Chinese (Simplified) to English:\nChinese (Simplified): 我爱机器翻译\nEnglish:"
inputs = tokenizer(text, add_special_tokens=False, return_tensors="pt")

outputs = model.generate(**inputs, max_new_tokens=1024)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Citation

@misc{han2026referencefreeposttrainingopenlarge,
      title={Reference-Free Post-Training of Open Large Language Models for Multilingual Machine Translation}, 
      author={Chris Han and Pengzhi Gao and Pei Fu and Jian Luan},
      year={2026},
      eprint={2608.10812},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2608.10812}, 
}

Limitations

MiLMMT-46 currently supports only the 46 languages listed above, and strong translation performance is not guaranteed for other languages. We will continue to improve the translation quality of MiLMMT-46, and future model releases will follow in due course.