Recursive ships Flint, a free coding model built largely by its own AI

Recursive, an AI startup founded last year, released its first model on October 21: Flint, a free coding model that runs on a laptop and, by the company's account, was built largely with the help of AI.

The work was done by Recursive's own automated research system, the Eureka Machine. According to the company, Eureka contributed to nearly every step from pre-training through post-training to inference, letting a small team build the model from scratch in under four months.

Architecturally, Flint is a mixture-of-experts model with 40 billion parameters in total, of which only 2.5 billion are active for any given task, which is what makes it light enough for an ordinary laptop. It is free to download on Hugging Face under the Apache 2.0 license, runs offline, keeps code local and carries no per-token usage fee.

Speed is the pitch. In Recursive's tests, Flint generated about 131 tokens per second on a MacBook Pro with Apple's M4 Pro chip, roughly twice the 58 to 69 tokens per second managed by comparable models such as Gemma 4 and Nemotron 3.5 Lightning on the same machine, and above 150 tokens per second on Nvidia's RTX Spark.

On benchmarks, Recursive says Flint beat every model under 50 billion parameters on SWE-Bench Verified, and that on its own six-test average only Alibaba's Qwen3.8 27B scored higher. All of these figures are vendor-reported and have yet to be independently reproduced.

The company concedes Flint needs more guidance than the big paid coding assistants and suggests breaking large tasks into smaller steps. Wired's Will Knight, testing it on a Mac with 48GB of memory, found that small coding models often error out or slow down, but that Flint nailed his coding requests every time.

Eureka's concrete contributions include a sparser model shape, low-bit training that shrank Flint by 71% with almost no quality loss, and custom kernels that speed up the underlying computation. Those are the visible engineering fingerprints of AI improving AI.

The significance is less about how strong Flint is than about what it demonstrates: a public, inspectable test of recursive self-improvement. If each model makes Eureka a better researcher, each generation should compress the next development cycle — which is precisely the most contested idea in the field, and one Recursive frames as human-supervised and interruptible.