Prime Intellect 发布 Prime Sandboxes：面向大规模智能体 RL 训练的 MicroVM
Announcements
Prime Intellect 推出 Prime Sandboxes，用 MicroVM 支撑大规模智能体强化学习训练。该产品于 9 月 23 日发布，属于其“开放超级智能栈”系列公告之一。同期该公司还公布了 1.3 亿美元 A 轮融资，并发布过 100B+ MoE 模型 INTELLECT-3。
Blog
Latest Updates fromPrime Intellect.
Prime Sandboxes: MicroVMs for Agentic RL Training at Scale
AnnouncementsJUL 08TH, 2026 $130M Series A to Build the Open Superintelligence Stack AnnouncementsMAY 28TH, 2026 Releasing Hosted Evaluations: Making benchmarking effortless AnnouncementsMAY 07TH, 2026 Releasing Lab: the training platform for self-improving agents AnnouncementsFEB 10TH, 2026 Introducing Lab: The Full-Stack Platform for Training your Own Models AnnouncementsNOV 26TH, 2025 INTELLECT-3: A 100B+ MoE trained with large-scale RL AnnouncementsOCT 27TH, 2025 Scaling Our Open-Source Environments Program AnnouncementsAUG 27TH, 2025 Environments Hub: A Community Hub To Scale RL To Open AGI AnnouncementsJUL 10TH, 2025 SYNTHETIC-2 Release: Four Million Collaboratively Generated Reasoning Traces AnnouncementsJUN 23RD, 2025 SYNTHETIC-2 AnnouncementsMAY 20TH, 2025 PCCL - Prime Collective Communications Library AnnouncementsMAY 11TH, 2025 INTELLECT-2 Release: The First 32B Parameter Model Trained Through Globally Distributed Reinforcement Learning AnnouncementsAPR 28TH, 2025 Planetary-Scale Inference: Building a Distributed Inference Engine for the Public Internet AnnouncementsAPR 15TH, 2025 INTELLECT-2: The First Globally Distributed Reinforcement Learning Training of a 32B Parameter Model AnnouncementsFEB 28TH, 2025 $15M to Build The Open Superintelligence Stack AnnouncementsFEB 20TH, 2025 SYNTHETIC-1 Release: Two Million Collaboratively Generated Reasoning Traces from Deepseek-R1 AnnouncementsFEB 06TH, 2025 SYNTHETIC-1: Scaling Distributed Synthetic Data Generation for Verified Reasoning AnnouncementsJAN 28TH, 2025 TOPLOC: A Locality Sensitive Hashing Scheme for Trustless Verifiable Inference AnnouncementsJAN 21ST, 2025 Distributed Training in the Inference-Time-Compute Paradigm AnnouncementsJAN 06TH, 2025 METAGENE-1: Metagenomic Foundation Model AnnouncementsNOV 29TH, 2024 INTELLECT-1 Release The First Globally Trained 10B Parameter Model AnnouncementsOCT 11TH, 2024 INTELLECT–1: Launching the First Globally-Distributed Training of a 10B Parameter Model AnnouncementsJUL 11TH, 2024 OpenDiLoCo: An Open-Source Framework for Globally Distributed Low-Communication Training AnnouncementsJUL 01ST, 2024 Introducing Prime Intellect Compute: The compute exchange AnnouncementsAPR 23RD, 2024 State-of-the-Art in Globally-Distributed Training
.png%26w%3D3840%26q%3D75&mode=full&exp=1791158400&sig=03629d56fd1e999f)
-min.png%26w%3D3840%26q%3D75&mode=full&exp=1791158400&sig=1c353861ddaca48a)
%252520(1)-1.png%26w%3D3840%26q%3D75&mode=full&exp=1791158400&sig=2ce42a72bca3e08e)
.png%26w%3D3840%26q%3D75&mode=full&exp=1791158400&sig=9e96bfbf549ea894)
来源：Prime Intellect（网页） · primeintellect.ai