Recently, MiniMax and Tencent Cloud have collaboratively carried out a significant initiative in the realm of Agent infrastructure. By harnessing the capabilities of Tencent Cloud, the two entities have successfully deployed an Agent Reinforcement Learning (RL) sandbox that boasts a throughput capacity in the millions and can handle concurrent operations in the hundreds of thousands. This sandbox functions seamlessly at full capacity within test environments, empowering MiniMax's reinforcement learning framework, Forge, to swiftly set up environments and promptly delete them post-use in large-scale Agent training scenarios. Such efficiency not only streamlines the training process but also enhances its stability and cost-effectiveness.
