Articles tagged "GENIAC"

A comprehensive analysis of Rakuten AI 3.0's Mixture of Experts architecture with 700B parameters. Explore the 8-expert configuration, 40B active parameter efficiency, and technical background behind achieving 8.88 on Japanese MT-Bench.

Rakuten AI 3.0, born from the GENIAC project led by METI and NEDO. Featuring a 700-billion-parameter MoE architecture and achieving an 8.88 score on the Japanese MT-Bench. Released under Apache 2.0 license, this article examines the strategic importance of domestic LLMs in Japan's AI industrial policy and the significance of data sovereignty.