Pinned Loading

1. arle arle Public

   Pure-Rust LLM runtime: one binary serves (OpenAI-compatible), runs local agents, and distills models on their own rollouts — on Apple Silicon and NVIDIA. No Python on the hot path.

   Rust 14 1

2. kernel-set kernel-set Public

   High-performance CUDA kernels for LLM inference & training — callable from Python, Rust, Go, and TypeScript through one stable C ABI.

   Python 2

3. eli eli Public

   Ease Lives Instantly

   Rust 5

4. openMax openMax Public

   The more, the better.

   Python 4