Miles v0.1: Open-Source RL Framework for LLMs Launches

Scobleizer · x · 2026-08-19

RadixArk released Miles v0.1, an open-source framework designed to simplify running and debugging reinforcement learning workloads for LLMs and multimodal models. Featuring over 1,326 commits and 85 GPU E2E CI tests, Miles has been battle-tested on frontier models like Kimi, DeepSeek, and Qwen, powering production RL workloads at companies including NebiusAI, Modal, and IBM. It supports both NVIDIA and AMD hardware, focusing on efficient hardware utilization and scalable execution.

Related event: RadixArk Open-Sources Miles v0.1, an RL Training Framework for LLMs(7 posts)→

Original post →

More from coding & agent

coding & agent channel →