CoreWeave Closes the Training-to-Inference Gap for Autonomous Agent Improvement

With the launch of unified agentic AI capabilities, enterprises can ship agents that improve using real-world data, paving the way for the superintelligence loop

LIVINGSTON, N.J.: CoreWeave, Inc. (Nasdaq: CRWV), The Essential Cloud for AI™, today announced the launch of unified agentic AI capabilities that accelerate progress toward the superintelligence loop, a closed feedback loop between training and inference. With reinforcement learning, production inference, agent observability, and autonomous improvement working as one closed loop, agents not only become more reliable, they compound in capability over time.

Until now, training reliable AI agents meant running lengthy offline evaluations for months before releasing them to real users for inference. Not only was this process too slow, but the agents often failed because the eval datasets couldn’t cover all possible real-world scenarios. As AI accelerates the path toward superintelligence, that process is no longer viable. CoreWeave eliminates this bottleneck, enabling enterprises to close the loop between training and inference. Now agents learn and improve as they operate in the real world.

Closing the Loop between Training and Inference

CoreWeave integrates four capabilities into a single closed loop:

• Training without the overhead: CoreWeave's Serverless RL enables enterprises to post-train large language models for reliability on multi-turn agentic tasks without provisioning or managing infrastructure. The service elastically scales with training workloads, reducing costs by up to 40% and accelerating training by approximately 1.4x with no loss in quality1. Training and inference run on separate always-on instances, so iteration cycles that previously took hours now take seconds.
• Inference built for production: CoreWeave Inference is designed to operate as a controllable, continuously running workload. This helps maintain reliable performance, runtime flexibility, and stable behavior under real-world traffic at scale. Built-in monitoring surfaces inference performance, scaling behavior, and system health, enabling teams to maintain production service level objectives as agent workloads grow.
• Visibility across every agent at scale: W&B Weave serves as the observability layer for the continuous loop between production behavior and agent improvement to achieve and maintain reliability. CoreWeave built new Weave capabilities from the ground up tailored specifically for agentic systems: production monitoring with built-in and custom signals that surface failure modes, a data model purpose-built for analyzing multi-agent workflows, and a flexible evaluation framework that prevents regressions as systems scale.
• Autonomous improvement: W&B Skills and MCP server turn general-purpose coding agents into AI researchers and agent builders that work around the clock to help create reliable agents autonomously. W&B Skills make coding agents instantly fluent in Weights & Biases’ leading AI tools for experiment tracking, model management, tracing, evaluations, and monitoring. The MCP server provides the tools and resources to access data and run experiments with Weights & Biases.

"The pace of AI has outrun the way teams build for it. Today's tradeoff: dev cycles that can't keep up, or shipping agents and discovering failure modes in production," said Chen Goldberg, Executive Vice President of Product and Engineering at CoreWeave. "Enterprises that put agents in production first and let them continuously improve from real-world experience aren't just building more reliable AI, they're accelerating the path to superintelligence."

“Most enterprises are stuck in a cycle of building and testing agents before they ever reach real users, and that cycle is becoming too slow and too expensive to sustain," said Nick Patience, Vice President & Practice Lead, AI Platforms, Futurum. "A platform that closes the production-to-development feedback loop, using real-world experience to automatically improve agent performance, addresses a critical bottleneck standing between enterprises and user-ready agentic AI. The teams that compress that iteration cycle will have a meaningful advantage over those that can't."

The Path to Reliable Agent Fleets

As AI agents take on increasingly complex, business-critical work, the ability to improve reliability, efficiency, and performance autonomously is becoming a defining competitive advantage. CoreWeave's unified agentic AI capabilities are designed to remove the barriers that have historically prevented enterprises from realizing that advantage at scale: fragmented tooling, GPU-intensive RL infrastructure, and the inability to translate production experience into systematic improvement.

The new CoreWeave capabilities are available now. Learn more here.

Built on proven AI infrastructure

CoreWeave consistently delivers industry-leading infrastructure performance, demonstrated by record-breaking MLPerf benchmark results, its position as the only AI cloud to earn the top Platinum ranking in both SemiAnalysis ClusterMAX™ 1.0 and 2.0, and its #1 ranking for inference speed and price-performance for Moonshot AI’s Kimi K2.6 in independent inference benchmarking conducted by Artificial Analysis.

About CoreWeave

CoreWeave is The Essential Cloud for AI™. Built for pioneers by pioneers, CoreWeave delivers a platform of technology, tools, and teams that enables innovators to move at the pace of innovation, building and scaling AI with confidence. Trusted by leading AI labs, startups, and global enterprises, CoreWeave serves as a force multiplier by combining superior infrastructure performance with deep technical expertise to accelerate breakthroughs. Established in 2017, CoreWeave completed its public listing on Nasdaq (CRWV) in March 2025. Learn more at www.coreweave.com.

_________________________
1 Compared to local H100 GPU environments.

Fonte: Business Wire