Prime Intellect Launches RL Sandboxes for Tens of Thousands of Concurrent VMs

Manual configuration of parallel environments often makes scaling AI training prohibitively expensive and complex. To address this, Prime Intellect has made its MicroVM sandboxes publicly available to facilitate the large-scale reinforcement learning (RL) required for agentic models. This infrastructure, previously an internal tool, provides full VM fidelity and elastic capacity and is accessible through a CLI/SDK or a dedicated RL suite. The platform uses a tier-free pricing model designed for high-volume workloads to lower the cost of maintaining laarge numbers of simultaneous environments. The company plans to expand the offering to include shared persistent workspaces, state snapshotting, and GPU microVMs to enable autonomous research loops.