Recursive Launches Flint, a Laptop Coding AI Model Developed Largely by AI

Recursive Introduces Flint AI Model

Recursive, an AI startup founded last year, has launched Flint, a free coding model designed to run on laptops. The company states that Flint was largely developed with the assistance of its automated research system, the Eureka Machine, enabling a small team to build the model in under four months.

Recursive Self-Improvement in Practice

Flint serves as an early public demonstration of recursive self-improvement, where AI contributes to building better AI. Recursive's approach involves an AI research assistant overseen by human developers, resulting in a model compact enough for laptop deployment. The company emerged from stealth with $650 million in funding from investors including GV, Nvidia, and AMD Ventures.

Offline Functionality and Cost Benefits

A key feature of Flint is its ability to operate directly on a user's computer, eliminating the need for cloud servers. This allows developers to work offline, maintain code privacy, and avoid usage fees associated with cloud-based AI services. Flint is available for free download on Hugging Face under the Apache 2.0 open-source license.

Technical Specifications and Performance

Flint is a mixture-of-experts model with 40 billion parameters, utilizing only 2.5 billion for any given task. This design makes it suitable for ordinary laptops, distinguishing it from larger models like Meta’s Muse Spark Open, which requires significantly more memory. In Recursive's tests, Flint generated approximately 131 tokens per second on a MacBook Pro with an M4 Pro chip, outperforming similar models like Google’s Gemma 4 and Nvidia’s Nemotron 3.5 Lightning on the same hardware. On Nvidia’s RTX Spark, it reached over 150 tokens per second.

On the SWE-Bench Verified benchmark, which assesses AI's ability to fix software bugs, Flint showed stronger performance than other models under 50 billion parameters.