Fast Technology reported on September 7 that Tencent Hunyuan announced that its Hunyuan translation model Hunyuan-MT-7B has topped the Hugging Face model trend list.
The official statement indicated that this model, along with the latest member of the Hunyuan world model family, HunyuanWorld-Voyager, occupies two of the top three positions.
It is reported that Hunyuan-MT-7B was open-sourced on September 1, with a total parameter count of only 7 billion, supporting 33 languages, and 5 types of Chinese dialect translations, making it a comprehensive lightweight translation model.
In the International Conference on Computational Linguistics (ACL) WMT2025 competition that concluded at the end of August, Hunyuan-MT-7B (competition name: Shy-hunyuan-MT) secured first place in 30 out of 31 language competitions, demonstrating an absolute lead.
These 31 languages include not only common languages like Chinese, English, and Japanese, but also less common languages such as Czech, Marathi, Estonian, and Icelandic.
Tencent Hunyuan stated that on the commonly used translation capability evaluation dataset Flores200, the Hunyuan-MT-7B model also showed excellent performance, significantly outperforming models of the same size and competing favorably with larger models.
For translation scenarios, Tencent Hunyuan proposed a complete training paradigm for translation models, covering the entire chain from pre-training, to CPT, then to supervised tuning, translation reinforcement, and integration reinforcement, enabling the model’s translation performance to reach industry-leading levels.
返回搜狐，查看更多