Recursive has released its first coding model, Flint, a free open-source system built largely with the help of an automated research architecture called the Eureka Machine. Founded last year and led by CEO Richard Socher, the startup developed the laptop-compatible model from scratch in under four months using its automated research system for nearly every step of the process.
Flint operates as a mixture-of-experts model with 40 billion total parameters, activating just 2.5 billion at a time to run efficiently on local hardware like a MacBook Pro. Available to download on Hugging Face under the Apache 2.0 license, the model enables offline development, data privacy, and zero cloud usage fees. In benchmark testing and early reviews, Flint demonstrated high processing speeds and strong performance on software bug-fixing tasks compared to other sub-50-billion-parameter models.
This release represents an early public test of recursive self-improvement in AI development, showcasing how automated research systems can accelerate the creation of future software.