AI BriefingPFNet Releases PLaMo 3 Translate Weights in Four Sizes on Hugging Face
·2026.11.07 14:43
Key point
The smallest 0.8B model outperforms existing translation-specialized LLMs with over 30 times the parameters in Japanese-English and multilingual tasks.
Details
- PFNet released PLaMo 3 Translate weights on Hugging Face in four sizes: 0.8B, 2B, 8B, and 31B.
- The 0.8B model outperforms existing translation-specialized LLMs, including those with more than 30 times as many parameters, in Japanese-English and Japanese-to-other-language translation.
- Performance gains were achieved through tournament-based data synthesis, back-translation, iterative self-distillation, and distillation into smaller models.
- Models support 53 languages, accept context and glossaries, and allow selection between plain and polite Japanese.
- Weights are available in BF16, FP8, W4A16, W4AFP8, NVFP4, GGUF, and MLX formats, with quantized versions retaining nearly the same accuracy as BF16.
- Released under the PLaMo Community License; commercial use is free for entities with revenue of JPY 1 billion or less after registration, while higher-revenue entities require a separate license agreement.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Thinking Machines AI introduces Inkling 2 Small, planning Apache 2.0 open weights release next week