“NEC's LLMs Set New Speed Standards, Surpassing GPT-4”
NEC Corporation has announced the creation of two large language models (LLMs) that it claims to outperform global LLMs, including OpenAI’s GPT-4 and Anthropic’s Claude 2, not just in veracity but more significantly, at over ten times their speed.
The Japanese tech company reported that its new “NEC cotomi Pro” model delivers comparable performance to these giants, while boasting of a response time 87% faster when using two Nvidia L40 GPUs.
On the other hand, its speedier “NEC cotomi Light” model matches up to global models such as “GPT-3.5-Turbo,” and even surpassed GPT-3.5 and GPT-4 in test conditions with its enhanced response rate and performance.
NEC attributed its triumphant improvement in both categories to pioneering training methods and architecture alongside a large Japanese dictionary's integration in the model. The specifics of these architectural innovations remain undisclosed.
Breaking down the effectiveness of these LLMs, “NEC cotomi Pro” produced results over five times faster than GPT-4 on a standard server with two GPUs, outpacing models like “Gemini 1.0 Pro” and comparing favorably to Claude 2 and GPT-4. Meanwhile, “NEC cotomi Light” outshone GPT-4 by producing results a staggering 15 times faster.
NEC intends on extending its development of these models and strengthen collaborations with partners to provide high-quality problem-solving AI services on the expanded NEC cotomi lineup.
Could this be the dawn of a new age redefining AI speed standards? Let us know your thoughts.
Also, don't forget to check out The Synthesia AI Video Creation Platform, for creating professional videos without mics, cameras, actors or studios by visiting www.TheBestAI.org/claim.
#NEC #AI #GPT4 #Technology
