Miles v0.1 Open Source RL Framework Launches for LLMs
ying11231 · x · 2026-08-19
SGLang/RadixArk launched Miles v0.1, an open-source reinforcement learning framework for LLMs and multimodal models. It addresses common challenges in RL training such as debugging difficulties, hardware efficiency, and scaling issues. Battle-tested over nine months with 1,326 commits and 85 GPU E2E CI tests, Miles supports frontier open models like Kimi K3, DeepSeek V4, and Qwen 3.8. It currently powers production RL workloads at companies including IBM, Modal, and DecagonAI on both NVIDIA and AMD hardware.
More from Infra
- 16-bit Model Requires 60GB VRAM; 4-bit Quantized Fits on 24GB — LeviTurk · 2026-08-19
- Why Removing the Vision Encoder Can Be Better — From an Infra Perspective — liuziwei7 · 2026-08-19
- Matmul Optimization Bottleneck: Data Movement, Not Multiplications — yaroslavvb · 2026-08-19
- Alibaba Cloud Opens 3rd Data Center in Korea; Doubao Adds PC Control — 创业邦 · 2026-08-19
- China opens world's largest AI data center targeting 1 million GPUs — teortaxesTex · 2026-08-19
- Morgan Stanley: US data centers need 68GW power by 2028 — tctjr · 2026-08-19