Hy-MT2-7B translation model
tencent/hy-mt2-7bTencent's open-weight 7B translation model for 33 language pairs and five Chinese dialect or minority-language pairs.
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 19 Aug 2026 · checked 3 Oct 2026.
About it
Hy-MT2-7B is Tencent's 7-billion-parameter translation model. It handles 33 language pairs plus five Chinese dialect and minority-language pairs. Its workflows include structured, delimiter-based, contextual, glossary-based, and style-guided translation. The model accepts and produces text, has an 8,192-token context window, and is available as open weights.
Review
In a line
Hy-MT2-7B is a text translation model made by Tencent. It has 7 billion parameters and covers 33 language pairs, along with five Chinese dialect and minority-language pairs.
What it's good at
The model is suited to translation workflows that need structure, delimiters, context, glossaries, or a chosen style. Those options can help when a translation must follow a set format or terminology list.
It works with text only.
Price in plain terms
Input costs $0.074 per million tokens, and output costs $0.295 per million tokens. Tokens are the small pieces of text that models process.
Reading a 500-word page costs under 0.01 cents. A chat turn with about 1,000 input tokens and 500 output tokens costs about 0.022 cents. One thousand such turns cost about $0.22.
Context and limits
Hy-MT2-7B can take up to 8,192 tokens in one context, so that is the total text it can consider at once. It can produce up to 4,096 tokens in one answer.
Its listed input and output type is text. Translation coverage is defined as 33 language pairs plus five Chinese dialect and minority-language pairs.
Ways to use it
The model is offered as open weights, with the model identifier tencent/Hy-MT2-7B. People can use those weights in software that supports the model.
Verdict
Choose Hy-MT2-7B if you need a Tencent translation model with many specified language pairs and workflows for context, glossaries, formatting, or style. Its open weights may suit teams that want to run the model in compatible software.
Look elsewhere if you need images or other non-text input and output, or if your language pair is outside the listed coverage.
Verdict
Hy-MT2-7B fits translation projects that can work within its stated language coverage and text-only design. Its open weights give compatible software a way to use the model, while the 8,192-token context and 4,096-token output limits set clear boundaries for longer material. The token rates are low for the example workloads provided, including about $0.22 for 1,000 chat turns.
- Supports 33 language pairs
- Adds five Chinese dialect and minority-language pairs
- Offers glossary- and style-guided translation workflows
- Open weights are available
- Input costs $0.074 per million tokens
- Text input and output only
- Context is limited to 8,192 tokens
- Maximum output is 4,096 tokens
- Coverage is limited to the listed language pairs
- Pick it ifText translation across its listed language pairs; Workflows needing glossaries, structure, or style guidance; Teams using open model weights
- Skip it ifYou need image or other non-text inputs and outputs; Your language pair is outside the stated coverage; You need more than 8,192 context tokens or 4,096 output tokens
Pricing
Input is $0.074 per million tokens and output is $0.295 per million tokens. The supplied examples put a 500-word page at under 0.01 cents and 1,000 sample chat turns at about $0.22.
Questions
What is Hy-MT2-7B?
It is Tencent's 7-billion-parameter text translation model.
How many languages does it support?
It supports 33 language pairs and five Chinese dialect and minority-language pairs.
What translation workflows does it support?
Its listed workflows are structured, delimiter-based, contextual, glossary-based, and style-guided translation.
How much does it cost?
Input costs $0.074 per million tokens and output costs $0.295 per million tokens. A 500-word page costs under 0.01 cents to read.
How much text can it handle at once?
Its context window is 8,192 tokens, and its maximum output is 4,096 tokens.
Can I use the model weights?
Yes. Hy-MT2-7B is listed as open weights, under the identifier tencent/Hy-MT2-7B.