Tencent: Hy-MT2-7B

Tencent: Hy-MT2-7B is a translation model from Tencent. It is a 7B-parameter model, which places it in the smaller end of the parameter range for models of this kind, and it sits in the text category. Its provider description frames it specifically as a translation system rather than a general purpose chat model, so the intended job is converting text from one language into another.

The model takes text as input and returns text as output. It has a context window of 8192 tokens, and the maximum output is 4096 tokens. According to Tencent, it supports 33 language pairs plus five Chinese dialect and minority-language pairs. Beyond plain sentence translation, it offers workflows for structured translation, delimiter-based translation, contextual translation, glossary-based translation, and style-guided translation. Those workflows matter because they let you control terminology and tone rather than accepting whatever the model produces by default.

The practical uses follow from that feature set. Glossary-based translation suits documentation, product catalogues and legal or technical material where specific terms must map consistently across languages. Contextual translation helps with longer passages where a sentence cannot be translated well in isolation, such as dialogue, support threads or multi-paragraph articles. Style-guided translation is useful when the same content has to read differently for different audiences, for example formal versus casual marketing copy. Delimiter-based and structured workflows fit pipelines where translated segments must be separated or aligned with their source, which is common in localization tooling. The Chinese dialect and minority-language coverage extends this to content aimed at those audiences.

You call the model through api14's OpenAI-compatible API using the model id "tencent/hy-mt2-7b". Because the interface follows the OpenAI chat completions shape, you can point an existing OpenAI client at api14 and change the model string. The supported request parameters are max_completion_tokens, max_tokens, response_format, stop, structured_outputs and temperature. Temperature gives you control over how literal or varied the output is, stop lets you cut generation at a chosen sequence, and response_format together with structured_outputs is useful when you need the translated result returned in a defined shape for downstream parsing.

The 8192-token context window is the main constraint to keep in mind. Very long documents will need to be split into chunks, and chunking interacts with translation quality because context is lost at the boundaries. The 4096-token output cap also means a single request cannot return an arbitrarily long translation. If your task is general reasoning, summarization or code rather than translation, a general purpose text model will usually be a better fit. For translation work within these size limits, this model is the more targeted choice.