Recursive releases Flint, a local coding model

Recursive's 21 October 2026 post introduces Flint, a mixture-of-experts model for agentic coding on local hardware. Recursive says it has 40 billion parameters in total and activates 2.5 billion per token, and that a small team built it from scratch in under four months with the company's Eureka Machine research system.

Recursive says Flint leads mixture-of-experts models under 50 billion parameters on SWE-Bench Pro v2, Terminal-Bench 2.1, and SWE-Bench Verified, and that it rivals dense models of 25 to 40 billion parameters while running more than five times faster. With Recursive's inference engine, the post says Flint generates more than 125 tokens per second on an M4 Pro MacBook Pro and more than 150 tokens per second on a 64GB NVIDIA RTX Spark at short contexts. Context goes up to 512,000 tokens. Recursive recommends staying within 256,000 for coding. The post says complex tasks need smaller subtasks and more explicit instructions than a larger closed model.

Weights are on Hugging Face under the Apache 2.0 license. Recursive says Flint works with llama.cpp, Ollama, Unsloth, LM Studio, vLLM, and SGLang, and that it is also shipping rec, a lightweight engine for M4 Pro and M5 Pro and Max machines. The post says the NVFP4 checkpoint scores within 0.7 points of the 16-bit model on Recursive's benchmarks.

Who it's for: someone who wants a coding model on their own machine, offline, without an API bill.

Fits: Best for Vibe Coding stays Cursor. Flint is a local model you run yourself. It sits beside the editor where you reject hunks. The verdict is unchanged.