Hy-MT2: A Family of Fast, Efficient and Powerful Multilingual Translation Models in the Wild

TL;DR

Hy-MT2 is a family of multilingual translation models optimized for speed and efficiency, supporting 33 languages and excelling in real-world translation tasks across various domains.

Contribution

Introduces a new family of multilingual translation models with multiple sizes, optimized for deployment and outperforming existing open-source and commercial models.

Findings

The 1.8B model requires only 440 MB storage with 1.5x faster inference.

7B and 30B models outperform open-source models like DeepSeek-V4-Pro.

The lightweight 1.8B model surpasses mainstream commercial APIs in performance.

Abstract

Hy-MT2 is a family of fast-thinking multilingual translation models designed for complex real-world scenarios. It includes three model sizes: 1.8B, 7B, and 30B-A3B (MoE), all of which support translation among 33 languages and effectively follow translation instructions in multiple languages. For on-device deployment, with AngelSlim 1.25-bit extreme quantization, the 1.8B model requires only 440 MB of storage and improves inference speed by 1.5x. Multi-dimensional evaluations show that Hy-MT2 delivers outstanding performance across general, real-world business, domain-specific, and instruction-following translation tasks. The 7B and 30B models outperform open-source models such as DeepSeek-V4-Pro and Kimi K2.6 in fast-thinking mode, while the lightweight 1.8B model also surpasses mainstream commercial APIs from providers such as Microsoft and Doubao overall.