Daily AI CatchupRecursive's Flint Model Codes Locally, Fast
Recursive Flint is a mixture-of-experts model built for agentic coding on local hardware. It has 40 billion total parameters but only activates 2.5 billion per token, letting it run at over 125 tokens per second on an M4 Pro MacBook and over 150 tokens per second on a 64GB RTX Spark. It was built from scratch by a small team in under four months and lets developers code without sending data to a remote server.
Why it matters
💻 Developer · A genuinely usable local coding model with real speed numbers on consumer hardware, worth trying if you want offline agentic coding.
📦 Product · Local-first AI coding tools reduce data residency concerns for teams that can't send code to third-party servers.
🎨 Design · Not directly design-relevant, but it shows small teams can now ship frontier-adjacent tools fast with AI-assisted development.
📈 Business · A small team shipped a competitive model in under four months, a data point on how fast AI development cycles are compressing.
🤔 Just Curious · This AI coding assistant runs entirely on your own laptop, no internet or company servers needed.
Sources: Introducing Flint