from Hacker News

Why small language models are the next big thing in AI

by dollar on 4/12/24, 8:27 PM with 1 comments

• by politelemon on 4/12/24, 9:34 PM

  > Like other SLMs, Gemma models can run on various everyday devices, like smartphones, tablets or laptops, without needing special hardware or extensive optimization.

  Is this as advertised, or slightly exaggerated? I had a look at the CodeGemma and Gemma and Cerule models on HuggingFace and the downloads are between 2 to 5 GB in size. FWICT this will still have a significant compute requirement which may not make it feasible to run on any smartphone.