SambaStack™

Yes. SambaStack supports both on-premises deployment and dedicated cloud hosting. Organizations can stand up their own AI data center using SambaStack hardware and have it operational within weeks, enabling them to process millions of tokens per day inside their own private environment.

SambaStack delivers fast inference on leading open-source models including Meta's Llama, DeepSeek, and gpt-oss-120b. The platform supports pre-configured model bundles that can be hot-swapped at inference time, allowing teams to serve multiple models without expanding their hardware footprint.

Yes. SambaStack is designed specifically for enterprise-grade security requirements. Organizations retain full control over their data. No prompts or outputs are shared with SambaNova or any third party, making it suitable for regulated industries and sensitive workloads.

Hot-swapping allows SambaStack to switch between different model configurations at inference time without taking the system offline. This means a single SambaRack can serve multiple AI models sequentially or on-demand, reducing the number of physical racks required to support a broad model portfolio.