It offers a "DeepTranslate" functionality in which the model reasons before making the translation. It results in much better results, as it takes context and idiomatic expressions into account.

In the future, tone (formal, informal, and custom), as well as infinite context and file translations, will be supported.

Note that even if the model reasons, it does not use a reasoning model, like o1 or o3-mini! It's a manual reasoning process performed with gpt-4o-mini, which has been shown to be the best translator besides Gemini 2.0 Flash.