About SambaNova
SambaNova sells AI inference on its own chips and rack systems. It offers a platform for running large models and sells hardware for data centers.
What it does
The hardware and software aim at fast, efficient model serving, including workloads for AI agents. Customers can use hosted inference or deploy SambaRack systems in their own facilities. The company also works with neocloud providers that resell AI capacity to their own customers.
Key features
- Inference on SambaRack SN50 systems
- Disaggregated inference for agent workloads
- Sizing tool for data center deployments
- Fine-tuning services for models
- Deployments for neocloud providers
Who it's for
- Enterprises planning dedicated AI infrastructure
- Neocloud providers that sell AI capacity
- Developers who want to test large models on new hardware
Pricing and good to know
Pricing for hardware and hosted inference is set through sales conversations. The company presents cost and payback figures in its own materials, so treat them as claims to verify with a quote for your workload. Check the developer resources to see whether any self-serve access is offered.
Highlights
- Run inference on SambaRack SN50 systems
- Serve agent workloads with disaggregated inference
- Size data center deployments with a tool
- Fine-tune models for specific tasks
