Foresight News reports that the distributed AI laboratory Gradient has released Echo-2, a distributed reinforcement learning framework designed to break through the barriers of AI research training efficiency. The framework decouples Learner and Actor at the architecture level to reduce the post-training costs of large models. According to official data, this framework can reduce the post-training cost of a 30B model from $4,500 to $425. Echo-2 uses compute-storage separation technology for asynchronous training (Async RL), supporting the offloading of sampling computation to unstable GPU instances and heterogeneous GPUs based on Parallax. The framework incorporates techniques such as bounded staleness, fault-tolerant scheduling for instances, and a self-developed Lattica communication protocol to improve training efficiency while maintaining model accuracy. Additionally, Gradient plans to launch the RLaaS (Reinforcement Learning as a Service) platform Logits, which is currently open for reservations to students and researchers.