Key Info

Recursive SI has released Flint, its first model, a ~40B parameter MoE open model designed for fast, accurate agentic coding on-device, and it is available on Ollama today.

Highlights

  • Flint is built with Recursive SI's automated research system, the Eureka Machine, and was built from scratch in under 4 months.
  • Reported throughput: 125+ tokens/sec on an M4 Pro MacBook Pro and 150+ tokens/sec on a 64GB NVIDIA RTX Spark on short contexts.
  • Recursive SI claims up to 2x faster performance than comparable models for fast local coding.
  • Run it with ollama run flint, or use it with Claude Code via ollama launch claude --model flint.