Recursive launches Flint, a laptop coding model built largely with AI

AI startup Recursive has released Flint, a free coding model designed to run on a laptop. The company says its automated research system, the Eureka Machine, helped build nearly every part of the model.

Recursive says a small team built Flint from scratch in under four months with the system’s help. The company emerged from stealth in May with $650 million in funding at a $4.65 billion valuation, backed by GV, Nvidia and AMD Ventures. It is led by CEO Richard Socher, Salesforce’s former chief scientist.

Flint is a mixture-of-experts model with 40 billion parameters, of which it uses 2.5 billion for a task at a time. In Recursive’s tests, it generated about 131 tokens per second on a MacBook Pro with Apple’s M4 Pro chip, compared with 58 to 69 tokens per second for similar models on the same machine. It exceeded 150 tokens per second on Nvidia’s RTX Spark. Recursive also says Flint outperformed other models under 50 billion parameters on SWE-Bench Verified.

The model is available to download for free on Hugging Face under the Apache 2.0 license. Because it runs locally, developers can use it offline and keep code on their own computers. Recursive says Eureka contributed techniques including low-bit training that reduced Flint’s size by 71% with almost no quality loss.

Flint is an early public test of recursive self-improvement, the idea that AI can help build better AI. Recursive describes Eureka as a research assistant overseen by people, who can stop the process if something goes wrong. The company did not say how much time the system saved.