Build predictable pricing.
Compare packages, configurations, infrastructure profiles, and regions.
On-Demand
One click, cost effective
Deployment model
- Deploy instantly any model across chosen geography.
- Automated GPU selection across clouds, regions, and instance types to optimize cost, availability, and price-performance.
- Transparent infrastructure-based pricing — $/hour. No token-based or cached-token pricing.
- Optimize costs with fractional GPUs, scale down to zero, and model-aware hardware recommendation.
- Broad catalog of supported open-weight models and GPU types.
Fully managed operations
Always-on- Self-serve provisioning and management.
- Telemetry, alerts, and usage metering.
- Built-in monitoring, health checks, and recovery.
- Automated upgrades and CVE patching.
- Managed infrastructure lifecycle and day-2 operations.
Standard Support
- Business-hours support.
- SLA-backed availability for production use.
Scale
Dedicated infrastructure and deployment flexibility
Deployment model
- Dedicated-GPU placement with predictable isolation and pricing.
- Run multi-GPU inference for higher-throughput workloads.
- Choose where and how your GPU workload runs.
- GPU arbitrage across clouds, regions, and instance types to optimize for cost, availability, and price-performance.
- On-demand scaling for dynamic workloads to save cost.
Fully managed operations
Always-on- Everything in On-Demand tier fully managed operations.
- High availability controls.
Priority Support
- Direct access to experts for performance, scaling, reliability, and runtime issues.
- Priority support through a private support channel.
Enterprise
Enterprise-ready for multicloud and hybrid workloads
Deployment model
- Run inside your private VPC to keep data and inference traffic within your perimeter.
- Leverage enterprise cloud commits without moving workloads outside your account.
- Deploy across on-prem, private cloud, and hybrid environments.
- Custom LoRA adapters for fine-tuning your models.
- Performance tuning insights with advanced GPU optimization controls.
Fully managed operations
Always-on- Everything in Scale tier fully managed operations.
- Built for enterprise security, data control, and operational governance — SOC 2 Type II and ISO 27001 certified, and GDPR and HIPAA compliant.
- Enterprise-grade availability.
Enterprise Support
- 24×7 critical incident coverage.
- Custom SLA and dedicated support.
- Hands-on architecture and operations support.
Networking included, subject to fair use. High-volume data transfer may require custom pricing.