Documentation menu

DocsDeployment options

Deployment options.

A Bridge deployment is three separate decisions: the model strategy, where private hardware lives, and which surfaces people use. Keeping them separate keeps the boundaries honest.

Decision 1 — model strategy

  • Cloud-only uses hosted cloud models from the providers you choose. It requires no dedicated private-inference server — the best fit when broad model access and minimal local infrastructure matter most.
  • Private-only runs models on dedicated, physical, GPU-capable hardware selected for your workloads and data boundaries. Private-only inference is not processed by cloud model providers.
  • Hybrid uses private or local models and cloud models together, routing each workload to the model that fits its sensitivity, capability, and operating needs.
Candid caveat

Hybrid describes model strategy. It does not mean sensitive content is automatically safe for every route, it does not mean Bridge Mobile access, and it does not describe who owns the hardware.

Decision 2 — private-hardware placement

This decision applies to private-only and hybrid strategies:

  • Customer-hosted. A customer-owned GPU server on-premises or in a location you control. You provide, own, and operate the hardware; Alluqi configures the software around the agreed environment.
  • Alluqi Managed. A dedicated, single-tenant, isolated private-model server hosted, maintained, and operated by Alluqi for your organization, under a custom monthly agreement with no public rate.
On-prem, defined

On-prem means the physical inference hardware is installed in your environment or another location you control. An Alluqi Managed server is hosted outside that environment — traffic leaves your network to reach it — so it is not on-prem.

Decision 3 — access surface

Bridge Desktop is the primary workspace. Bridge Mobile adds optional paired reach to active sessions. Your interface decision does not change the model strategy and does not move the server.

Side by side

DecisionCloud-onlyCustomer-hosted privateAlluqi Managed privateHybrid
Model locationProvider cloudYour hardwareDedicated Alluqi-hosted serverBoth, routed per workload
Hardware ownerThe providerYouAlluqiDepends on the route
Data boundaryProvider termsYour networkSingle-tenant isolated server outside your networkPer route
Best fitBroad model access, minimal infrastructureStrict on-prem boundaryPrivate models without operating hardwareMixed-sensitivity workloads
See the full deployment guide