Recursive released Flint, a coding model it claims was built largely by AI, on Wednesday. The startup, which specialises in systems that improve themselves, made the tool available for free on a laptop.
How the model was made
The company’s automated research system, the Eureka Machine, handled nearly every stage of Flint’s creation. This allowed a small team to build the model from scratch in under four months. The Eureka Machine generates ideas, runs experiments, and decides the next steps based on results. It contributed to every part of the build, from pre-training through to inference.
Recursive says the system helped create a sparser model shape and low-bit training that reduced the file size by 71% with almost no quality loss. It also produced custom kernels, specialised programs that speed up the underlying calculations.
Performance and specs
Flint is a mixture-of-experts model. It uses only a small slice of itself for any given task, activating just 2.5 billion of its 40 billion parameters at a time. This makes it light enough for an ordinary laptop.
In Recursive’s tests, the model generated about 131 tokens per second on a MacBook Pro with an Apple M4 Pro chip. That is roughly twice the 58 to 69 tokens per second that similar models managed on the same machine. Those competitors include Google’s Gemma 4 and Nvidia’s Nemotron 3.5 Lightning. On Nvidia’s RTX Spark, Flint topped 150 tokens per second.
On SWE-Bench Verified, a benchmark that measures how well AI can fix real-world software bugs, Flint outperformed other models under 50 billion parameters. Across six tests, only one model, Alibaba’s Qwen3.8 27B, scored higher.
For context, Meta’s Muse Spark Open is an open model built for Macs that was released last week. It uses 816 billion parameters and needs at least a terabyte of memory.
What it means for developers
Because Flint runs on a user’s own computer, developers can use it offline. They can keep their code private and avoid the usage fees that cloud services charge based on text processed. The model is free to download on Hugging Face under the Apache 2.0 license, which allows free use and modification.
Early reviews were positive. Wired’s Will Knight used Flint on a Mac with 48GB of memory to build a to-do list app and a tool that scours the web for research. He found that smaller coding models often ran into errors or were slow compared to Claude or Codex, but Flint nailed his coding requests every time.
The company acknowledged that Flint would need more guidance than the big paid coding assistants. It suggested breaking large tasks into smaller steps.
Background and future plans
Recursive emerged from stealth in May with $650 million in funding at a $4.65 billion valuation. Backers include GV, Nvidia, and AMD Ventures. The company is led by CEO Richard Socher, Salesforce’s former chief scientist.
Recursive previously told TechCrunch that its $410 million compute deal with Amazon Web Services in June was “less about headcount and more about agent count.” OpenAI said last month that by mid-August, its AI agents were logging 3.1 workdays for every one worked by a human researcher.
The longer-term goal is for the Eureka Machine to achieve full recursive self-improvement. Humans would oversee the process and be able to stop it if something goes wrong. Recursive says each new model should make the Eureka Machine a better researcher, helping it build the next one faster.
