Skip to main content
Flint is Recursive’s model for fast, local agentic coding, built with our autoresearch system, the Eureka Machine. Run it locally or in the cloud and connect it to your coding harness to get started.

Join our Discord community

Meet other developers using Flint, ask questions, and share feedback with the Recursive community.

​
Local usage

Run Flint locally with rec, Ollama, Unsloth, LM Studio, or llama.cpp. Choose an option below for setup instructions. For a quick start on a supported Mac, we recommend rec with Pi. rec downloads the model and connects the harness automatically.

rec

Run Flint on your Mac with Recursive’s optimized inference engine and automatic setup for Pi or OpenCode.

Ollama

Download and run Flint with Ollama, then connect it to your coding harness.

Unsloth

Use Flint in Unsloth’s local app and connect it to your coding harness.

LM Studio

Import Flint for desktop chat and local serving with a compatible runtime.

llama.cpp

Run Flint with a compatible GGUF and runtime build, with direct control over GPU offload and local serving.

​
Coding harnesses

Connect Flint to a coding harness using a local runtime or cloud provider.

OpenCode

Connect OpenCode to a local Flint runtime, or use hosted Flint-40B on its free tier, subject to capacity and rate limits.

Pi

Try Flint free through Radius, or launch locally with rec.

Cline

Use Flint in Cline Desktop, VS Code, or the terminal through a local model server.

OpenHands

Connect OpenHands to a local Flint runtime or hosted provider.

​
More ways to run Flint

For model behavior and recommended defaults, see the Flint overview and generation settings. For specifications and checkpoints, see the model card.

​
A first task worth trying

Start with a small repository you know. Ask Flint to explain one function, add a focused test, then fix the behavior that test exposes. Check the diff and test output after each step. For setup issues, see troubleshooting. For downloads and release information, see release status.