CoreWeave launches full-stack AI cloud platform CoreWeave Forge with RL Rollouts
CoreWeave Inc. has introduced CoreWeave Forge, a full-stack AI cloud platform designed to streamline the AI improvement loop for enterprises. The platform includes RL Rollouts, a reinforcement learning capability currently in preview, and cross-region write support for its AI Object Storage. These tools aim to enhance GPU utilization and accelerate post-training workflows, as demonstrated by a collaboration with You.com to post-train Nvidia’s Nemotron 3.5 Lightning model in eight hours.
Editor, Lazyfounder

30 SEC SUMMARY
- CoreWeave Inc. has launched CoreWeave Forge, a full-stack AI cloud platform designed to optimize GPU utilization during AI post-training and model improvement.
- The platform introduces RL Rollouts, a reinforcement learning capability currently in preview, to support repeated cycles of model refinement.
- CoreWeave also announced cross-region write support for its AI Object Storage, enabling faster post-training data sharing.
- Partner You.com used RL Rollouts to post-train Nvidia’s Nemotron 3.5 Lightning model in just eight hours.
TABLE OF CONTENTS
KEY HIGHLIGHTS
- CoreWeave unveiled CoreWeave Forge, a full-stack AI cloud platform aimed at accelerating the AI improvement loop for enterprises.
- RL Rollouts, a reinforcement learning capability, is now in preview to support continuous model improvement cycles.
- CoreWeave AI Object Storage now supports cross-region writes, improving post-training data accessibility.
- You.com and CoreWeave post-trained Nvidia’s Nemotron 3.5 Lightning model in eight hours using RL Rollouts and web search tools.
- CoreWeave Forge integrates tools like Agent Lens, Registry, RL Rollouts, and ARIA to streamline AI workflows.
CoreWeave Forge aims to optimize AI post-training
CoreWeave Inc. has launched CoreWeave Forge, a full-stack AI cloud platform designed to improve GPU utilization during AI post-training. According to SiliconANGLE, the platform connects model deployment, evaluation, and improvement, addressing inefficiencies in data movement and model updates.
The platform introduces RL Rollouts, a reinforcement learning capability currently in preview. This tool supports repeated cycles of generating training responses and updating models, enabling faster iteration and refinement.
Cross-region data writes enhance post-training efficiency
CoreWeave AI Object Storage now supports cross-region writes, allowing post-training jobs to write results back for immediate use by other teams. This feature aims to reduce delays in collaborative AI workflows, particularly for distributed teams.
The capability was demonstrated in a collaboration with You.com Inc., which used RL Rollouts and its web search tools to post-train Nvidia’s Nemotron 3.5 Lightning model in just eight hours. This partnership highlights the potential for faster model improvement cycles using CoreWeave’s infrastructure.
CoreWeave advocates for an open AI ecosystem
CoreWeave is positioning itself as a proponent of an open, full-stack AI cloud ecosystem that supports multiple models, frameworks, and clouds. The company’s platform is built specifically for AI workloads, unlike traditional infrastructure designed for earlier web applications.
The broader platform integrates compute, networking, storage, software, and operations to support AI workloads in production. According to SiliconANGLE, CoreWeave’s approach addresses gaps in speed, performance, and tooling that clients often encounter when migrating from other clouds.
Platform tools streamline the AI improvement loop
CoreWeave Forge aims to accelerate the AI improvement loop, which consists of five stages: run, observe, curate, improve, evaluate, and repeat. The platform integrates tools like Agent Lens for inspecting agent activity, Registry for managing checkpoints, RL Rollouts for model improvement, and ARIA (AI Research and Iteration Agent) to help users identify patterns.
The company emphasized that the speed of learning is the most critical metric in this loop. By providing a shared workflow for business units and machine learning teams, CoreWeave Forge aims to reduce friction and accelerate AI development.
CoreWeave expands partner network for faster AI adoption
CoreWeave has launched a partner network featuring tested, co-engineered integrations to provide enterprises with playbooks and use cases for faster value realization. You.com Inc. has joined this network, contributing web search tools for AI agents.
The company’s platform has already attracted enterprise customers, including Cognition AI Inc., which is running production workloads on CoreWeave’s Nvidia Vera Rubin NVL72 racks. CoreWeave was also the first to validate Nvidia’s Vera Rubin platform and has earned SemiAnalysis’ Platinum ClusterMAX rating three times.
What this means
Lazyfounder analysis — our interpretation, not reported fact.
CoreWeave Forge represents a shift toward full-stack AI platforms that prioritize not just raw compute power but also the tools and workflows needed to improve models efficiently. For founders and operators, this signals a growing need to evaluate AI infrastructure beyond GPU access. The ability to accelerate the AI improvement loop—through reinforcement learning, cross-region data sharing, and integrated tooling—could be a differentiator for startups aiming to iterate quickly and deploy models at scale.
The partnership with You.com also underscores the value of collaboration in the AI ecosystem. Startups should consider how specialized partners can enhance their workflows, whether through web search tools, storage solutions, or other complementary technologies. CoreWeave’s focus on an open ecosystem suggests that interoperability and flexibility will remain critical as AI workloads grow more complex.
Key takeaways
- Founders building AI models should evaluate full-stack platforms like CoreWeave Forge to optimize GPU utilization and reduce post-training inefficiencies.
- Reinforcement learning tools like RL Rollouts can significantly accelerate model improvement cycles, reducing time-to-market for AI applications.
- Cross-region data capabilities in AI storage solutions can enhance collaboration and speed up iteration for distributed teams.
- Partnering with specialized AI infrastructure providers can help startups achieve faster and more efficient model training and deployment.
Related on Lazyfounder
Sources
- SiliconANGLE · 2026-10-06
CoreWeave targets GPU utilization in continuous AI post-training - SiliconANGLE · 2026-10-08
CoreWeave makes the case for an open, full-stack AI cloud - SiliconANGLE · 2026-10-08
CoreWeave Forge aims to speed up the AI improvement loop
This story is an original summary drafted with AI by Lazyfounder from the reporting listed above and checked by automated validation. Facts are attributed to their original publishers; sections marked as analysis are Lazyfounder's. Where a source is in another language, facts were machine-translated and quotations are reported, not reproduced. Read the original coverage via the links, and see our AI policy and corrections policy.
About the author
Editor, Lazyfounder
Tarun Mottlia edits LazyFounders, covering Indian startups, funding rounds, AI and product launches. Every story on the site is AI-assisted and checked against its cited sources before publication.
More stories by Tarun MottliaGet the LazyFounder Brief
Startup, funding and AI news in a five-minute read. Join the early-access list.


