Hy-MT2-1.8B-GGUF

Hy-MT2-1.8B-GGUF

Translates text among 33 languages and obeys translation instructions, as a 1.8B model in GGUF format for llama.cpp. For developers who want a small local translator.

Made by tencent on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: Hy-MT2-1.8B-GGUF is the smallest model of Tencent's Hy-MT2 translation family, packaged as a GGUF file for llama.cpp. It covers 33 languages and obeys translation instructions given in several of them. The authors say it beats commercial translation services from Microsoft and Doubao overall.

Who it suits: It suits developers who want a small translator they can run themselves, and teams that translate content or subtitles between the supported languages.

Industry | Who uses it | How they use it | Example
Localization | Translation teams | Draft translations between the supported languages | Turning a Chinese product description into English
Software | App developers | Add a small local translator to an app | A translate button that runs on the user's machine
Media | Subtitle teams | Translate subtitle lines with an instruction on tone | Rendering a short video's subtitles in another language

You can skip it if you need the larger 7B or 30B-A3B models, or a model for general chat, because this family is built for translation.

Where to use it:

• llama.cpp, which this GGUF version targets
• Windows, macOS or Linux, with Python for the download step
• Agent setups, through the Hy-MT2-Translator Skill that Tencent released for these models

Use case | Who | What they get back
Text translation | Developers | A translation between two of the 33 languages
Instructed translation | Localization teams | A translation that follows a written instruction
Local translator | App builders | A small model that runs in llama.cpp

How to set it up:

• Python, to run the Hugging Face download tool
• The huggingface_hub package
• A llama.cpp setup that can load GGUF files
• Room for about 896 MB of model file, as a 4-bit estimate

1. Install the Hugging Face hub package.
2. Download the repository files to your computer.
3. Load the gguf file in llama.cpp.

pip install -U huggingface_hub
hf download tencent/Hy-MT2-1.8B-GGUF

How to use it: You give the model your text and say in plain words which language you want. The part of the card shown here has no run command, so follow llama.cpp's own steps for loading a GGUF file.

Tool | What it does
hf download | Fetches the model files to your computer
llama.cpp | Loads and runs the GGUF file
Hy-MT2-Translator Skill | The skill Tencent released for using Hy-MT2 models in translation work
IFMTBench | A benchmark Tencent released for translation instruction following

Example: You ask, "Translate this paragraph into Chinese and keep the tone formal." The model returns the paragraph in Chinese, following the tone instruction you gave.

For newcomers and experts: If you are new, a translation model turns text from one language into another. To start, download the files and open the gguf file in llama.cpp with a short sentence.

If you know the field, Hy-MT2 comes in 1.8B, 7B and 30B-A3B sizes, with FP8, 2-bit and 1.25-bit variants of the small one. The card says 1.25-bit quantization brings the 1.8B model down to 440 MB and speeds inference by 1.5x. That is a separate download.

You give text and a language instruction
llama.cpp loads the GGUF file
The 1.8B model translates the text
You get the translation back

Good to know:

Title: Model sizes in the Hy-MT2 family
Smallest: 1.8 billion
Middle: 7 billion
Largest: 30 billion

• Languages: 33
• Size: about 896 MB at 4-bit as an estimate
• Other variants: FP8, 2-bit and 1.25-bit builds, plus the 7B and 30B-A3B models
• Benchmark claims come from Tencent's own evaluation
• License: apache-2.0
• Cost to run: none named
• Data it reaches: the text you give it, and the files you download from Hugging Face
• Not for: general chat, since the family is built for translation

On Hugging Face, 13 November 2026

• Parameters: 1.8B
• Weights: about 896 MB at 4-bit (estimate)
• Formats: GGUF
• Downloads in the last 30 days: 382,300
• Likes: 257
• License: apache-2.0
• Based on: tencent/Hy-MT2-1.8B
• Quantized versions: 2
• Fine-tunes: 1
• Tags: gguf, conversational

Links

• Model: https://huggingface.co/tencent/Hy-MT2-1.8B-GGUF
• Files: https://huggingface.co/tencent/Hy-MT2-1.8B-GGUF/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download tencent/Hy-MT2-1.8B-GGUF

Then follow the usage in its model card. Nothing in this kit runs by itself.