Japan's Preferred Networks releases the weights of its PLaMo 3 Translate models in four sizes, from 0.8B to 31B parameters

Japanese AI company Preferred Networks (PFN) has put the weights of PLaMo 3 Translate, its translation-focused language model, on Hugging Face in four sizes: 0.8B, 2B, 8B and 31B parameters, each with several compressed versions that run locally on ordinary computers and Macs. The models translate between 53 languages, accept surrounding context and glossaries, and let users pick plain or polite Japanese. PFN says the 0.8B model beats other translation-focused models more than 30x its size on Japanese-English and Japanese-to-other-language translation. Personal and noncommercial use is free; companies with revenue of JPY 1 billion or less last fiscal year can use it commercially for free after registering, and larger ones need a separate license.

Why it matters

All four sizes are downloadable, the smallest runs offline, and smaller companies can use them commercially for free, giving teams that want to run translation themselves another option.

Sources1

Written by dalao.ai: the title, summary and why it matters are our own words. The original reports are the record; their publishers keep authorship.