Skip to content
Configurations

Three starting points for on-premise AI.

Kernel, Tower and Rack are planning categories. Capacity, models, integrations, timing and commercial terms are confirmed from the institution's measured requirements.

How to choose a Hosn configuration

Kernel, Tower and Rack are planning categories for three deployment scales. They do not represent fixed public hardware bundles. The right starting point depends on measured concurrency, corpus size, model memory, latency, availability, integrations and facility constraints.

QuestionKernelTowerRack
Starting scopeDepartmentMultiple teamsInstitution-wide
ArchitectureSingle applianceMulti-nodeRack-scale GPU infrastructure
CapacityMeasured and confirmed during discovery
ModelSelected and validated for the workload
Commercial termsWritten quotation and contract

Readiness assessment

The readiness assessment defines the workload, users, corpus, data classification, network boundary, directory and logging integrations, facility constraints and acceptance tests. Its output is a written scope that can be reviewed before a quotation is accepted.

Model selection

Hosn evaluates model families against representative, approved material. Model quality is workload-specific, so an exact variant is not promised from a generic page. The selected model, runtime and context setting are documented in the delivery scope.

Pricing and availability

Every configuration is priced by written quotation. No public price, discount, delivery date or maintenance percentage applies universally. The quotation and contract are the authority for the commercial scope.

Configuration guides

Next step. Start with a scope discussion involving the institution's IT and data owners. The readiness assessment documents the workload, data boundary, integrations and delivery requirements before a written quotation is prepared.

Book a briefing