We're open-sourcing Hy-MT1.5-1.8B-1.25bit — a 440MB translation model that runs fully offline on you...
TL;DR · AI 摘要
腾讯混元开源 Hy-MT1.5-1.8B-1.25bit 翻译模型：仅440MB，支持33种语言+5种方言，1.25-bit量化无损精度，手机端全离线运行，性能超越Google Translate及部分商用API。
核心要点
- 25-bit超低比特量化实现440MB体积，较FP16压缩7.5倍且零精度损失
- 支持1056个翻译方向，覆盖藏语、蒙古语等少数民族语言
- 在标准机器翻译基准上媲美235B参数大模型和商业API
结构提纲
按章节快速跳转。
- §发布概览
宣布开源Hy-MT1.5-1.8B-1.25bit翻译模型及其核心指标。
通过创新量化方案将模型从3.3GB压缩至440MB，速度提升10%，精度无损。
支持33种语言、5种方言、1056个翻译方向，含藏语、蒙古语等稀缺语种。
在主流MT基准上超越Google Translate，已落地多个腾讯产品并获30项国际竞赛冠军。
提供Android APK、Hugging Face模型页、GitHub代码库及论文链接。
思维导图
用一张图看清主题之间的关系。
查看大纲文本（无障碍 / 无 JS 友好）
- Hy-MT1.5-1.8B-1.25bit
- 核心技术
- 1.25-bit量化（无损）
- 1.8B参数规模
- 能力特性
- 33语言+5方言+1056方向
- 离线手机端运行（440MB）
- 实证表现
- 超越Google Translate
- 媲美235B模型/商用API
- 30项国际MT竞赛冠军
金句 / Highlights
值得收藏与分享的关键句。
By quantizing to 1.25-bit, memory drops from 3.3GB (FP16) to 440MB — 25% smaller and ~10% faster than prior 1.67-bit approaches, with no accuracy loss.
Covers 33 languages, 5 dialects, and 1,056 translation directions including minority languages like Tibetan and Mongolian.
At 1.8B parameters, it matches commercial translation APIs and 235B-scale models on standard benchmarks.
Our translation model has won 30 first-place rankings in international MT competitions and is already deployed across multiple Tencent products.
At 1.8B parameters, it matches commercial translation APIs and 235B-scale models on standard benchmarks. By https://t.co/4kgWldn4p1" / X
We're open-sourcing Hy-MT1.5-1.8B-1.25bit — a 440MB translation model that runs fully offline on your phone, supports 33 languages, and outperforms Google Translate. At 1.8B parameters, it matches commercial translation APIs and 235B-scale models on standard benchmarks. By quantizing to 1.25-bit, memory drops from 3.3GB (FP16) to 440MB — 25% smaller and ~10% faster than prior 1.67-bit approaches, with no accuracy loss. Covers 33 languages, 5 dialects, and 1,056 translation directions including minority languages like Tibetan and Mongolian. Our translation model has won 30 first-place rankings in international MT competitions and is already deployed across multiple Tencent products.
Demo APK (Android): huggingface.co/AngelSlim/Hy-M
Hugging Face:: huggingface.co/AngelSlim/Hy-M
GitHub: github.com/tencent/AngelS
Paper: arxiv.org/abs/2601.07892
The media could not be played.