Liquid AI introduces Hyena Edge, a convolution-based LLM optimized for smartphones, achieving 30% faster prefill and decode times compared to Transformers; tested on Samsung Galaxy S24 Ultra, it uses less RAM for improved edge device efficiency

Liquid AI unveils “Hyena Edge,” a convolution-based LLM designed for smartphones, offering 30% faster prefill and decode latencies than Transformer models. Tested on a Samsung Galaxy S24 Ultra, Hyena Edge uses less RAM, improving efficiency on edge devices.

Source: venturebeat.com