from Hacker News

Phi-3 14B model with 128k context

by Handy-Man on 5/21/24, 3:24 PM with 1 comments

  • by causal on 5/21/24, 4:48 PM

    Really impressive for the size. The 4.2B vision model apparently outperforms GPT-4V-Turbo on 5 out of 7 vision benchmarks.