Miles v0.1 Runs Fully Asynchronous Agentic RL on a 744B GLM-5.2 at 263 Seconds Per Step

arXiv 2609.08368·medium signalXBlueskyLinkedInCopy link

Miles v0.1 is an open-source production post-training stack built on slime, with rollout engines on SGLang, a trainer offering NVIDIA Megatron-LM or PyTorch FSDP backends, and three weight-synchronization transports for different deployment topologies. Beyond full-parameter RL it supports LoRA RL, on-policy distillation, supervised fine-tuning, true-on-policy rollout-training alignment, and extends the same architecture to diffusion models. The end-to-end case study runs fully asynchronous agentic RL on a GLM-5.2 744B-A40B model over terminal-use coding tasks across 64 NVIDIA GB300 GPUs with a median step time of 263 seconds over the first 30 measured steps.