Miles system builds reliable scalable training for large AI models
Miles v0.1: Production-Level Post-Training
Machine LearningComputation and Language
Summary
Training very large AI models can be complicated and slow. The authors introduce Miles, a complete system designed to make training smarter and bigger AI models easier and more efficient. It focuses on clean, customizable parts that work well together, supporting different training methods and hardware setups. They show how it works by training a very large AI model for coding tasks on many GPUs efficiently. The system is freely available for anyone to use and build upon.
What this means in practice
- •For ai engineering teams: Use Miles to build and run efficient and scalable reinforcement learning training pipelines for large AI models in production environments.
- •For cloud infrastructure teams: Deploy and manage high-performance AI training systems with flexible hardware support and weight synchronization protocols across GPU clusters.
Authors
RadixArk, :, Tom Chen, Mao Cheng, Shi Dong, Kangrui Du, Yanbin Jiang, Jiajun Li, Yiming Li, Tao Lin, Yusheng Su, Andy Ye, Yueming Yuan, Zhichen Zeng
Abstract
We present Miles v0.1, a full-stack, production-ready system for frontier post-training. Building upon the clean design of slime, Miles designs each stage of the reinforcement-learning (RL) training loop around a single principle: components should be verified, clean, and customizable. With accuracy, efficiency, reliability, and scalability as first-class goals, Miles aims to make frontier-scale RL accessible to researchers and enterprises alike. This report walks through the system end to end: rollout engines built on SGLang, a trainer with a choice of two backends (NVIDIA Megatron-LM and PyTorch FSDP), and three weight-synchronization transports for different deployment topologies. Beyond full-parameter RL, Miles also supports LoRA RL, on-policy distillation, supervised fine-tuning, and true-on-policy rollout-training alignment, and extends the same architecture to diffusion models. We close with an end-to-end case study: fully asynchronous agentic RL on a GLM-5.2 744B-A40B model over terminal-use coding tasks, running on 64 NVIDIA GB300 GPUs with a median step time of 263 seconds over the first 30 measured steps. Miles is open-sourced at https://github.com/radixark/miles, with the project website at https://miles.radixark.com.