Best models for English → Japanese and English → Chinese translation?

Hi everyone I’m working on machine translation and need support for both:

• English → Japanese

• English → Chinese

I wanted to understand best practice here :

• Is it better to use separate language‑pair models (e.g., en‑ja and en‑zh individually)?

• Or is there a single multilingual model that performs well for both (like NLLB or MADLAD) without a noticeable drop in quality?