A look at Meta's next generation Open Source LLM

What's Shown

This graphic shows Meta's newest release of LlaMa 3 as compared to previous iterations, using parameter count as the main metric.

Why LLaMa?

Zuckerberg's approach to AI is at odds with the rest of the industry. His goal is to disseminate (very expensive to train) models out to as many groups, corporations and individuals as possible, for free, with full, under-the-hood access. This philosophy is fundamentally different than more safety-focused firms publishing powerful, yet completely secretive, models of their own. With his newest unreleased model, the 405B parameter LLaMa 3, he looks to fundamentally change the way users interact with their products, as well as the internet at large.