Recursive, the AI startup led by former Salesforce chief scientist Richard Socher, has launched Flint — a coding model designed to run entirely on a laptop. The company says Flint is free and open source under an Apache 2.0 license, and claims it was built largely by its automated “Eureka Machine” research system, from pre-training through inference optimization.
Flint uses a mixture-of-experts design with 40 billion total parameters but only 2.5 billion active at a time — which is what lets it fit on consumer hardware. Recursive reports about 131 tokens per second on a MacBook Pro with an M4 Pro chip, roughly double the 58-69 tokens per second it measured for Google’s Gemma 4 and NVIDIA’s Nemotron 3.5 Lightning on the same machine. On an NVIDIA RTX Spark GPU, the company says Flint exceeds 150 tokens per second.
On SWE-Bench Verified, the benchmark for fixing real software bugs, Recursive says Flint outperformed all models under 50 billion parameters except Alibaba’s Qwen 3.8 27B. It also claims the Eureka system shrank the model by 71% with low-bit training and near-zero quality loss — a reminder that model compression is becoming as strategically important as raw scale. For developers comparing assistants, see our reviews of Cline, Codeium, and Supermaven, plus our explainer on quantization, the technique that makes models like Flint laptop-sized.
Source: Global1.News
