Skip to content

Latest commit

 

History

History
41 lines (25 loc) · 837 Bytes

File metadata and controls

41 lines (25 loc) · 837 Bytes
id category title
45
On_Device
Liquid AI 1.2B vs Google 4B

Liquid AI 1.2B vs Google 4B

A 1.2B model just beat Google's 4B model at Korean-English translation. Here's how ↓

𝗪𝗮𝗶𝘁, 𝘄𝗵𝗮𝘁?

CHrF++ 34.61 (vs Gemma-3 4B: 32.83) Trained in just 400 steps Quantizes to Q8 with epsilon-close-to-zero performance loss Runs locally under 1.25GB

This model is perfect for → Edge translation apps → Private document translation → Mobile-first products

Beautiful

𝗪𝗵𝗲𝗿𝗲'𝘀 𝘁𝗵𝗲 𝗰𝗼𝗱𝗲?

Full guide + training notebooks ↳ https://lnkd.in/dGhV3it8

𝗟𝗼𝗼𝗸𝗶𝗻𝗴 𝗳𝗼𝗿 𝗺𝗼𝗿𝗲 𝗵𝗮𝗻𝗱𝘀-𝗼𝗻 𝗲𝘅𝗮𝗺𝗽𝗹𝗲𝘀 𝗼𝗳 𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁 𝗔𝗜?

122

love 마음에 쏙듬