PrismML built the first ever commercial 1-bit LLM. Why does that matter? Most models store each weight as a 16-bit floating-point number; 1-bit models store them as simply +1 or –1. The takeaway that matters is that 1-bit models can be smaller, faster, and massively more eco-friendly with a relatively small loss in performance. In other words, the intelligence per gigabyte or per watt blows traditional models out of the park. Aside from the obvious environmental benefits, 1-bit models can run locally on edge devices like phones.
We worked with PrismML to design their visual identity, marketing site, and data visualizations, which are central to explaining advantages over competitors.