CoreWeave, the AI Hyperscaler™, unveiled three new AI cloud software products at the Weights & Biases Fully Connected Conference, marking the first major product integration since acquiring Weights & Biases in May 2025. The new capabilities aim to help AI developers and researchers iterate faster, monitor model performance in real time, and streamline large-scale deployments.
The launch includes:
-
Mission Control Integration: CoreWeave’s advanced cluster health management tool is now embedded within W&B Models, giving developers the ability to correlate infrastructure issues like node failures or hardware errors directly with training runs. This integration provides real-time diagnostics and actionable insights to reduce training downtime and improve reliability.
-
W&B Inference: Built on the CoreWeave Cloud Platform, this feature provides streamlined access to top open-source models—including DeepSeek R1-0528, LLaMA 4 Scout 17Bx16E, and Phi 4 Mini 3.8B—via a unified interface, simplifying the experimentation and deployment process.
-
W&B Weave Online Evaluations: This tool allows developers to monitor how AI agents perform in production across any cloud platform. It enables ongoing evaluation with custom datasets, ensuring continuous quality and rapid issue resolution.
Also Read: Smart Communications Debuts SmartHUB to Modernize Enterprise Archiving and Compliance
“Today’s announcement demonstrates how CoreWeave and Weights & Biases together can accelerate and enable our customers to unleash the potential of AI,” said Chen Goldberg, SVP of Engineering at CoreWeave.
Lukas Biewald, GM of Weights & Biases, added, “In this first joint launch, we have delivered for our customers with metal-to-token observability and inference integrated with application-level observability.”
These innovations signal CoreWeave’s deepening focus on delivering end-to-end performance, scalability, and transparency for AI professionals. The tools are now available in preview, giving early access to teams looking to enhance development workflows on one unified AI platform.