Apple’s latest AI breakthrough, OpenELM, is an open-source language model that significantly outperforms earlier models like OLMo by being 2.36 percent more accurate while using only half the pretraining tokens. This state-of-the-art model, developed using texts from diverse public sources, incorporates advanced techniques such as layer-wise scaling and RMSNorm to enhance performance and efficiency. With its open-source availability, OpenELM provides unprecedented access to its training processes and performance benchmarks, encouraging widespread innovation and collaboration in the AI research community.
It's Just Technology