May 29, 2025 - NVIDIA B200 instances available
NVIDIA B200 instances are now available on CoreWeave, providing the latest in AI compute performance
Update Instances NVIDIA HGX B200 instances are now Generally Available on CoreWeave, providing next-generation AI compute capabilities. These instances deliver exceptional performance for the most demanding AI workloads, from large language model training to real-time inference.
Overview
NVIDIA HGX B200 instances are now Generally Available on CoreWeave, providing next-generation AI compute capabilities. These instances deliver exceptional performance for the most demanding AI workloads, from large language model training to real-time inference.
Key features
| Feature | Description |
|---|---|
| NVIDIA HGX B200 | Latest generation GPU architecture optimized for AI workloads |
| Next-generation AI | Exceptional performance for large language models and AI applications |
| High-performance networking | InfiniBand connectivity for distributed training |
| Managed infrastructure | Fully managed by CoreWeave with automated scaling and monitoring |
| Production ready | Generally Available status ensures stability and support |
Performance characteristics
AI workload performance
- Large language models: Optimized for training and fine-tuning large language models
- Distributed training: Excellent performance for multi-GPU and multi-node training
- Real-time inference: Low-latency inference capabilities for production workloads
- Memory bandwidth: High memory bandwidth for memory-intensive AI operations
Compute capabilities
- Tensor cores: Advanced tensor operations for AI workloads
- Multi-precision support: Support for various precision formats (FP16, FP32, etc.)
- Scalability: Efficient scaling across multiple GPUs and nodes
- Energy efficiency: Optimized power consumption for cost-effective operations
Use cases
B200 instances are ideal for:
| Use case | Description |
|---|---|
| Large language model training | Training and fine-tuning large language models |
| Computer vision | Image recognition, object detection, and video processing |
| Natural language processing | Text analysis, translation, and generation |
| Recommendation systems | Large-scale recommendation and personalization |
| Scientific computing | Computational research and data analysis |
| Real-time inference | Production inference workloads with low latency |
Technical specifications
For detailed specifications of B200 instances, see B200 with InfiniBand.
Key specifications
- GPU architecture: NVIDIA HGX B200
- Memory: High-bandwidth memory optimized for AI workloads
- Networking: InfiniBand connectivity for distributed workloads
- Storage: High-performance storage options
- Operating systems: Support for major Linux distributions
Availability and pricing
General availability
B200 instances are now Generally Available, meaning:
- Production ready: Stable and supported for production workloads
- Full support: Complete technical support and documentation
- SLA guarantees: Service level agreements for reliability
- Long-term availability: Committed to long-term availability and support
Pricing
- Competitive pricing for next-generation AI compute
- Flexible pricing options for different usage patterns
- Volume discounts for large-scale deployments
- Contact your CoreWeave account team for detailed pricing
Getting started
Deployment options
- CoreWeave Cloud Console: Deploy through the web interface
- API deployment: Use CoreWeave APIs for automated deployment
- Terraform: Infrastructure as code deployment with Terraform provider
- Kubernetes: Deploy as part of CKS clusters
Migration considerations
When migrating to B200 instances:
- Application compatibility: Verify your applications are compatible with B200
- Performance testing: Test performance improvements in your specific use case
- Cost optimization: Evaluate cost vs. performance benefits
- Gradual migration: Consider gradual migration from existing instances
Documentation
For detailed information about B200 instances and deployment options, see:
Support and resources
Technical support
- 24/7 support: Round-the-clock technical support
- Expert assistance: AI and GPU specialists available
- Documentation: Comprehensive documentation and guides
- Community: Active community for knowledge sharing
Training and resources
- Best practices: AI workload optimization guides
- Performance tuning: Tips for maximizing B200 performance
- Case studies: Real-world examples and success stories
- Webinars: Regular training sessions and updates