Skip to content
Halo by OmniTrust

Optimize · Model Routing

Guided rollout· Route profiles and fallback in the model gateway

Match each workload to an approved model

Halo's model gateway supports configurable route profiles, model catalogs with pricing, and fallback paths. Cost-aware routing creates savings potential by matching each workload to an approved model that fits its cost, quality, and policy needs.

Book a walkthrough Start with Halo
configurable in the model gateway
Route profiles
included in model catalogs
Pricing

Why it matters

One default model decides your AI bill

Agents and apps often call whichever model a developer picked first, whether or not it fits the task. That can mean paying premium rates for routine work, or stalling when a provider has trouble.

  1. 01

    Premium models end up handling routine tasks that a lower-priced approved model could do.

  2. 02

    A provider outage or rate limit stops work when no fallback path exists.

  3. 03

    Switching models by hand can bypass the review that governance expects.

How it works

From model catalog to measured results

A guided rollout helps your team choose models and profiles before production work moves over.

  1. 1

    Build the catalog

    Start from the models your organization approves, from Halo-managed models or your own keys, with pricing attached.

  2. 2

    Define route profiles

    Create profiles that map kinds of work to approved models, and add fallback paths. Halo's team guides this setup.

  3. 3

    Route through the gateway

    Requests flow through the model gateway, which applies identity, policy, approvals, and budgets before it calls the selected model.

  4. 4

    Measure and refine

    Compare cost and usage by agent and model, then adjust profiles and fallbacks where the numbers point.

Capabilities

Routing inside the governed path

Routing lives in the model gateway, so routing choices sit next to policy, approvals, budgets, and evidence.

  • Route profiles

    Define route profiles in the model gateway that say which approved models serve which kinds of work.

  • Model catalogs with pricing

    Catalogs list the models available to your account along with their pricing, so routing choices can weigh cost.

  • Fallback paths

    Configure fallback paths so a request has an approved alternative when its primary route cannot serve it.

  • Cost-aware matching

    Match workloads to approved models that fit their cost, quality, and policy needs, which creates savings potential for routine work.

  • Governance stays in force

    Routing happens inside the model gateway, so identity, policy, approvals, budgets, and evidence still apply to the model that serves a request.

  • Cost visibility

    Cost analytics by agent and model show how routing changes affect spend, so you can adjust profiles with evidence in hand.

Savings potential

Cost-aware routing with limits you set

Cost-aware routing creates savings potential when routine work goes to an approved model with a lower price and demanding work keeps a stronger one. Policy and budgets still apply, and cost analytics show what actually changed.

  • Profiles choose among models you have approved
  • Catalog pricing informs the choice
  • Fallback paths give requests an alternative
  • Budgets and policy still apply to routed calls

FAQ

Questions, answered

Model routing sends a model request to an approved model according to a route profile. Profiles, model catalogs with pricing, and fallback paths are configured in the Halo model gateway, so routing decisions sit next to policy and budgets.

Halo does not promise a savings figure. Matching each workload to an approved model that fits its cost, quality, and policy needs creates savings potential, and cost analytics by agent and model show the effect so you can adjust.

No. Routing runs inside the model gateway, so identity, policy, approvals, budgets, and evidence apply to the model that serves a request. A route selects among models your organization has approved.

Route profiles and fallback in the model gateway are available through a guided rollout. Our team helps you choose models, define profiles, and set fallbacks before you move production workloads onto them.

Routes draw on Halo-managed models and on BYOK providers, which include OpenAI, Anthropic, Google Gemini, Cerebras, and GMI Cloud. More than 100 other providers are available through guided setup, so confirm your provider list with our team.

Related capabilities

Model GatewayManaged and BYOK model accessLearn more about Model GatewayBudgets and Cost ControlsBudgets that notify, pause, or stopLearn more about Budgets and Cost ControlsGovernance DashboardUsage, decisions, and cost in one viewLearn more about Governance Dashboard

Put a halo around the AI your business runs.

See discovery, runtime policy, approvals, and evidence working together on your own use cases.

Book a walkthrough Start with Halo