Recursive, a startup led by former Salesforce chief scientist Richard Socher, has released Flint, a coding model designed to run locally on laptops. The company utilized its automated research system, the Eureka Machine, to develop the model from scratch in under four months.
Flint is a mixture-of-experts model with 40 billion parameters, utilizing 2.5 billion parameters per task. It is available for free on Hugging Face under the Apache 2.0 license. The model allows for offline use, ensuring code privacy and eliminating cloud-based usage fees.
In performance tests on a MacBook Pro with an M4 Pro chip, Flint generated approximately 131 tokens per second, outperforming models like Google’s Gemma 4 and Nvidia’s Nemotron 3.5 Lightning. On Nvidia’s RTX Spark, the model exceeded 150 tokens per second.
The Eureka Machine contributed to the model's development, including low-bit training that reduced its size by 71% and the creation of custom kernels. Recursive previously raised $650 million at a valuation of $4.65 billion, with backing from GV, Nvidia, and AMD Ventures.
