PLaMo 3 Translate vs Hy-MT2: which open translation model?
PLaMo 3 Translate covers 53 languages under a revenue-limited license, while Hy-MT2 covers 33 under plain Apache 2.0 and has a far larger community. PFN's own charts favor PLaMo.
Verdict. Check the license and your language pair first. PLaMo 3 Translate (0.89B to 31.8B parameters, 53 languages) was tuned and measured mostly on Japanese-English and Japanese-to-other-language translation, and it takes a glossary, context and a plain or polite Japanese style as dedicated options. Its community license is free for personal and academic use, but commercial use needs registration with PFN, a limit of 1 billion yen of revenue and the PLaMo name in derived models. Hy-MT2 (1.8B, 7B and 30B-A3B, 33 languages plus Tibetan, Kazakh, Mongolian, Uyghur and Cantonese on its language table) is plain Apache 2.0, has templates that keep tags and structured data intact and a 256K context. PFN's own charts put every PLaMo size ahead of Hy-MT2, and Tencent's report does not test PLaMo, and we found no independent comparison. Translate a sample of your own text with both.
Specifications
| Specifications | PLaMo 3 Translate | Hy-MT2 |
|---|---|---|
| Developer | Preferred Networks | Tencent |
| Released | November 5, 2026 | May 21, 2026 |
| Total parameters | 31.8B (PLaMo 3 Translate 31B) | 30.1B (Hy-MT2-30B-A3B (MoE)) |
| Active parameters | All | 3B |
| Architecture | Dense | Dense |
| Context window | 32K tokens | 256K tokens |
| Input | Text | Text |
| License | PLaMo Community (Conditional) | Apache 2.0 (Permissive) |
| Memory for the smallest variant (4-bit) | 0.614 GB (PLaMo 3 Translate 0.8B) | 1.41 GB (Hy-MT2-1.8B) |
Performance claims come from the developer and have not been independently verified.
Pick PLaMo 3 Translate if
- Your main pair is Japanese and English, or Japanese and another language, which is what PFN tuned and measured.
- You need more languages than Hy-MT2's 33, because PLaMo lists 53.
- You want a glossary, context and plain or polite Japanese as dedicated options instead of prompt text.
- You want a 0.8B or 2B model for a laptop CPU, with GGUF files of 0.57 GB and 1.8 GB.
Pick Hy-MT2 if
- You need plain Apache 2.0, with no registration, revenue limit or naming rule.
- You want the larger community: Hugging Face showed about 1.16 million downloads in 30 days for Hy-MT2-7B-GGUF, against about 18,000 for PLaMo 3 Translate 8B GGUF.
- You translate text with tags, placeholders, JSON or HTML that must stay intact, which the Hy-MT2 card gives templates for.
- You translate Tibetan, Mongolian, Uyghur or Cantonese, which PLaMo's language list does not include, or long documents that need more than PLaMo's 32,768 tokens of input and output together.