Unverified50% confidenceFactExact time
GShard achieved a 600-billion-parameter multilingual translation model in 2020
1
Sources
50%
Confidence
Long-term
Relevance
8/15/2026
First Seen
Sources
Gemini Flash文档处理能力被低估？基准测试验证用户真实体验
Related Claims
Unverified2006年Google翻译初始系统使用了一个在数万亿词文本上训练的5-gram语言模型，这很可能是世界上第一次万亿token规模的语言模型训练64% similarUnverifiedMeta's NLLB (No Language Left Behind) model supports translation across more than 200 languages and is fully open-source.61% similarUnverifiedThe taxonomy proposed by Joshi et al. in 2020 classifies global languages into six resource levels from 0 to 5.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/749718API
curl https://kongchang.com/api/v1/knowledge/claims/749718MCP
get_claim(id=749718)