Recursive's Flint is a free coding model made to run on your own laptop.

Recursive, the AI startup led by former Salesforce chief scientist Richard Socher, released its first model, Flint, on Wednesday, Oct 21, with free open weights under the Apache 2.0 license (Recursive). It's a "mixture-of-experts" model. It has 40 billion parameters in total, but only about 2.5 billion do the work for each word it writes, which is why it can run quickly on a laptop. Recursive says it writes more than 125 tokens a second on an M4 Pro MacBook Pro and more than 150 on a 64 GB NVIDIA RTX Spark at short context lengths, and that it leads mixture-of-experts models under 50 billion parameters on coding tests like SWE-Bench Pro and Terminal-Bench. Those are Recursive's own numbers.

Recursive says its automated research system, the Eureka Machine, helped build nearly every part, from the training data and recipe to the code that runs it on each chip, and that a small team built it from scratch in under four months. WIRED's Will Knight tested it on a Mac with 48 GB of memory and wrote that "Flint nailed my coding requests every time" (WIRED).

Recursive's own advice is that it "may need more guidance than the larger closed-source coding models you're used to," so break big jobs into smaller steps.

Recursive has raised $650 million at a $4.65 billion valuation, according to WIRED.

Try it: if you have a Mac or PC with plenty of memory, install Ollama and run ollama run flint:40b. That download is about 23 GB.

It also works with LM Studio, llama.cpp and other common tools. The full-size version on Hugging Face is about 80 GB and meant for a data-center GPU. Your code stays on your machine and there are no API fees, so it's an easy test against the paid coding agent you use now.