Recursive Launches Flint, a 40-Billion-Parameter Language Model for Local Coding

Recursive, a company founded by former DeepMind researchers, has introduced Flint, its first language model developed through an AI-driven research process. According to The Decoder, Flint is a 40-billion-parameter mixture-of-experts (MoE) language model designed specifically for local coding tasks.

The model was created in just four months using Recursive's proprietary AI system, known as the Eureka Machine. This rapid development showcases the efficiency of Recursive's in-house AI research framework, as reported by The Decoder.

For Japanese markets, where AI applications in software development and automation are growing rapidly, Flint’s emergence signals potential advancements in coding productivity and localized AI solutions tailored to regional needs.