while Phi-3-small and Phi-3-medium outperform larger models like GPT-3.5 Turbo. Recently, Apple's DataComp for Language Models (DCLM) team released a new open-source model called DCLM-7B under the ...