Forge Free
Designed for personal development of AI applications and models.
- AI application evaluations
- AI application tracing
- AI application scorers
- AI model experiment tracking
- AI assets registry & lineage tracking
- Community Support
AI API Tool

PRODUCT OVERVIEW
CoreWeave is an AI cloud platform built for training and serving models at scale. It provides NVIDIA GPU compute, bare-metal Kubernetes, high-performance storage, and low-latency networking, plus CoreWeave Forge, a connected development layer for running, observing, curating, improving, and evaluating models and agents. Forge includes Weights & Biases Models, Agent Lens, Registry, Sandboxes, Notebooks, Training, Inference, and ARIA. The platform targets AI labs, enterprises, and research teams that need to move from experimentation to production. Access options include on-demand and spot GPU instances, Forge Free, Forge Pro, and Forge Enterprise, with academic research access available.
Create an account at id.coreweave.com/signup to start on Forge Free or begin a Forge Pro trial.
Use the framework-agnostic SDK to connect models and agents running on any cloud or on-prem infrastructure.
Run models and agents against real workloads, then use Agent Lens traces and monitors to capture behavior.
Flag failures into versioned datasets in Weights & Biases Models, then fine-tune, apply reinforcement learning, or distill with CoreWeave Training.
Score candidates against production traces, use ARIA to explain metric changes, and record passing checkpoints in Registry.
Training large-scale AI models
Running low-latency AI inference
Developing and evaluating AI agents
Fine-tuning open models
Reinforcement learning
Model distillation
Data preprocessing and pipelines
VFX rendering
CoreWeave Forge is the development layer of the CoreWeave platform, bringing together Weights & Biases Models, Agent Lens, Registry, Sandboxes, Notebooks, Training, Inference, ARIA, and Automations in one connected environment.
No. Forge connects to workloads wherever they run, including on-prem infrastructure and other cloud providers, and its SDK is framework and LLM agnostic.
Yes. You can prototype in Notebooks, run agent code in isolated Sandboxes, compare models and prompts in Playground, and use Agent Lens and Evaluations to monitor and score agents.
Yes. CoreWeave Training supports supervised fine-tuning, reinforcement learning, and model distillation with no training cluster to stand up.
Nothing changes. Your credentials, projects, and wandb.ai links carry over, and Weights & Biases Models is now part of CoreWeave Forge.
Forge Pro is designed for early-stage teams and organizations with fewer than 50 employees. Customers who exceed these guidelines will be required to transition to Forge Enterprise.
CoreWeave Forge provides a free Pro license to academic institutions pursuing research unrelated to a for-profit entity, intended for students, professors, and postdoctoral researchers with an active academic email address.
The Pro plan is billed monthly upfront. Model seats are prorated if added mid-term, while Weave data ingestion, serverless inference, sandboxes, and storage are billed monthly in arrears based on usage.
Prices are in USD as shown on CoreWeave's official pricing pages (checked 2026-11-19). Forge Pro starts at $60/month billed monthly and is limited to early-stage teams under 50 employees; larger customers must move to Enterprise. Forge Enterprise and some GPU models (GB300, HGX B300, RTX PRO 6000 Standard Memory) are custom/contact-sales. Forge ARIA is free only for a limited promotional period and will become consumption-based, so it is not listed as a free plan. Reserved compute offers up to 60% discounts over on-demand.
Designed for personal development of AI applications and models.
For professionals optimizing AI applications and models; starts at $60/month, billed monthly, with a 30-day free trial. For early-stage teams fewer than 50 employees; larger customers must transition to Enterprise.
For companies prioritizing security and compliance for AI applications and models; custom plans, contact us.
On-demand 8x NVIDIA HGX H100 instance, North America region.
On-demand 8x NVIDIA HGX H200 instance, North America region.
On-demand 8x NVIDIA A100 instance, North America region.
Hot tier of CoreWeave AI Object Storage; no ingress, egress, or transfer fees.
Warm tier of CoreWeave AI Object Storage; no ingress, egress, or transfer fees.
Cold tier of CoreWeave AI Object Storage; no ingress, egress, or transfer fees.
COMMUNITY REVIEWS
No reviews yet.