DeepSeek Elastic Compute (DSec) is a production sandbox platform designed for large-scale agentic training and evaluation with large language models (LLMs). It exposes Fn Call, container, micro VM, and full-VM sandbox backends through a unified SDK. DSec coordinates placement and lifecycle management across clusters, composes environments from independently versioned layers, and combines memory sharing, reclamation, and CPU scheduling for high-density execution.

It loads image data on demand from Fire-Flyer File System (3FS), a cluster-wide distributed filesystem. Co-designed with reinforcement learning (RL) frameworks, DSec decouples stateful rollout execution from preemptible GPU training, coordinates sandbox lifecycle with training to preserve rollout state while reclaiming idle resources, and mitigates agent misbehavior such as reward hacking. A single production-scale unit spans around 160 nodes, serving about 3 million sandboxes per day.

In production, it supports over 380,000 concurrent sandboxes and sustains over 5,000 sandbox creations per second. Evaluation shows reduced environment setup and image-distribution overhead, improved memory efficiency, and preserved latency-sensitive performance under high-density overcommit.