Hy-MT2-1.8B Translation Model Review
tencent/hy-mt2-1.8bA compact Tencent translation model with open weights, 33 language pairs, and text-only output.
8K tokens of context — about 6,144 words, on a scale that ends at 2M.
What it can do
- Open weights
- Takes text
- Gives text
Released 20 Aug 2026 · checked 3 Oct 2026.
About it
Hy-MT2-1.8B is Tencent’s 1.8-billion-parameter translation model. It handles 33 language pairs plus five Chinese dialect and minority-language pairs. The model supports structured, delimiter-based, contextual, glossary-based, and style-guided translation workflows. It uses text input and output, has open weights, and reads up to 8,192 tokens at a time.
Review
In a line
Hy-MT2-1.8B is a compact translation model made by Tencent. It works with text and has open weights.
What it's good at
The model covers 33 language pairs and five Chinese dialect and minority-language pairs. It is designed for translation workflows that can use context, glossaries, or style guidance.
It also supports structured and delimiter-based workflows, which can help organize translation input and output.
Price in plain terms
Input costs $0.044 per million tokens, and output costs $0.177 per million tokens. Tokens are the small pieces of text that models read and write.
Reading a page of about 500 words costs under 0.01 cents. A chat turn with about 1,000 tokens in and 500 out costs about 0.013 cents. One thousand such turns cost about $0.13.
Context and limits
Hy-MT2-1.8B can read up to 8,192 tokens in one request and write up to 4,096 tokens in one answer.
It handles text only. The facts identify translation coverage and workflow types, but do not list other capabilities or supported software.
Ways to use it
The model is available with open weights. Its listed Hugging Face identifier is tencent/Hy-MT2-1.8B, which can help identify it in compatible model tools.
The model is text in and text out. The facts do not name particular apps, chat tools, or service providers.
Verdict
Choose Hy-MT2-1.8B if you need a small, text-based translation model with broad language-pair coverage and open weights. Its glossary, context, and style-guided workflows suit translation tasks that need more control than plain sentence conversion.
Look elsewhere if you need image or audio handling, a context window above 8,192 tokens, answers longer than 4,096 tokens, or named app integrations.
Verdict
Hy-MT2-1.8B fits buyers who want low-cost text translation and the freedom to work with open model weights. Its coverage includes many language pairs and several Chinese dialect or minority-language pairs, while its workflow options support controlled translation. The model is less suitable for multimodal work, since it accepts and returns text only. It also may not fit projects that need very long documents in one request or a long single response. No specific apps or providers are identified here, so compatibility must be checked before adoption.
- Supports 33 language pairs and five Chinese dialect or minority-language pairs
- Open weights are available
- Context window reaches 8,192 tokens
- Input and output prices are low at the stated rates
- Text-only input and output
- Maximum output is 4,096 tokens
- No specific app integrations are listed
- Pick it ifLow-cost text translation; Projects needing open model weights; Glossary- or style-guided translation workflows
- Skip it ifYou need images, audio, or other non-text input and output; You need more than 8,192 tokens of context; You require named integrations with specific apps
Pricing
Input is $0.044 per million tokens and output is $0.177 per million tokens. A 500-word page costs under 0.01 cents to read; 1,000 stated chat turns cost about $0.13.
Questions
What is Hy-MT2-1.8B?
It is Tencent’s 1.8-billion-parameter translation model with open weights.
How many language pairs does it support?
It supports 33 language pairs, plus five Chinese dialect and minority-language pairs.
How much text can it read at once?
Its context window is 8,192 tokens.
How long can one answer be?
The maximum output is 4,096 tokens.
What does it cost?
Input costs $0.044 per million tokens; output costs $0.177 per million tokens.
Is it multimodal?
No. The listed modality is text, with text inputs and text outputs.
Where is the model identified?
Its listed Hugging Face identifier is tencent/Hy-MT2-1.8B.