SAN FRANCISCO, CA, Oct 7, 2026 – CoreWeave has released CoreWeave Forge, a development environment that connects AI model and agent training, production monitoring and evaluation. Its tools use production data to inform subsequent versions, with support for different models, frameworks and clouds.
MasterClass and Canva are already building on CoreWeave Forge. The news was shared during Fully Connected, CoreWeave’s AI cloud conference, which brings together more than 4,500 customers, partners, developers and AI leaders.
CoreWeave Forge combines Weights & Biases Models, post-training expertise from OpenPipe, the open-source marimo notebook project and CoreWeave services. It connects execution, observation, data curation, improvement and evaluation in a repeated development cycle. Training runs, experiment tracking, evaluations and agent traces share the environment where the model or agent runs.
“As more people build with AI, they’re putting models to work with their own data and workflows. That’s where the gaps between model capability and system performance become clear,” said Chen Goldberg, executive vice president of product and engineering at CoreWeave. “Engineering teams need to understand those gaps, identify the signals that matter, improve the next version and measure whether the change worked under real operating conditions. All of this has to function as a single, connected system. Forge brings all these capabilities on one platform, so what a business learns from running AI becomes part of how engineers improve it.”
Forge includes the following capabilities and editions:
CoreWeave Forge is available today with new and expanded capabilities across the AI loop:
- CoreWeave ARIA: Now available, ARIA is a coding agent that automatically analyzes experiment data and agent observability data. It identifies factors behind changes and recommends subsequent experiments and agents to run, with supporting evidence.
- Weights & Biases Models: Tracks and compares tens of thousands of experiments and millions of metrics using visualizations and automated workflows.
- CoreWeave Notebooks: Supports prototyping, training, evaluation and production in the same environment. Custom visualizations are built on marimo.
- CoreWeave Agent Lens: Analyzes tens of millions of production-agent traces to identify failures and inform fixes. CoreWeave reports that Agent Lens detects 20% more critical failures and fixes issues at one-tenth the cost of a general-purpose frontier large language model.
- CoreWeave Sandboxes: Now available, the service provides a fresh, isolated environment for each run involving agent tool use, reinforcement learning (RL) or model evaluation. It connects to the other Forge tools.
- CoreWeave Registry: Versions model checkpoints and agent configurations in open, portable formats with lineage information. Teams can resume work from earlier versions on different technology stacks.
- CoreWeave Post-Training: Includes serverless RL, serverless supervised fine-tuning and model distillation. CoreWeave reports that serverless RL trains 1.4 times faster at 40% lower cost than a self-managed setup.
- CoreWeave Inference: Part of CoreWeave Forge, the service provides playground and inference deployment access for open-weight models. Dedicated Inference also includes RL Rollouts in preview, which loads checkpoints into a live deployment so training can continue without redeployment.
- CoreWeave Forge Free, Pro and Enterprise editions: Forge is available in three editions.
“Every platform decision in this market has involved weighing how much optionality a team is willing to surrender for a connected toolchain,” said Nick Patience, vice president and practice lead, AI platforms at the Futurum Group. “CoreWeave Forge is built on the premise that teams shouldn’t have to make that calculation. Closing the loop from production back into training in a single environment, while staying open across models, frameworks, and clouds, is a harder engineering problem than either half alone.”
Source: CoreWeave
About CoreWeave
CoreWeave provides cloud infrastructure and managed services for artificial intelligence and high-performance computing workloads. The company operates GPU-based infrastructure, data centers, and software for training, inference, and large-scale AI deployment. Its services include cloud computing, Kubernetes management, storage, networking, and infrastructure orchestration. CoreWeave also develops software to manage GPU resources and computing infrastructure. Its customers include AI companies, cloud service providers, enterprises, research organizations, financial services firms, media companies, and software developers. Organizations use its infrastructure for AI model training, inference, visual effects, scientific computing, and other compute-intensive workloads. CoreWeave was founded in 2017 and is headquartered in Livingston, NJ. CoreWeave reported 2,100 employees worldwide. The company operates data centers in the United States and Europe.
