Small open source model just beat a giant one on my own test

I ran a side by side on my laptop last night, a 7B parameter model from Mistral against the big API I pay for monthly. The local one got my coding task right in one try, no extra tokens spent. Anyone else finding that smaller models are good enough for daily stuff now?

1 comments

Log in to join the discussion

Log In

1 Comment

aaron_ellis4219d ago

My 2019 MacBook runs a 3B model faster than my phone loads the big API.

1