Token Economics capability
Send each job to the right model. By your policy.
Routing rules make AI model choice a deliberate cost decision, not a fixed default.
- 1
intent = FAQ and prompt < 2k tokens→Small model - 2
needs multi-step reasoning→Large model - 3
team budget above 90% this month→Small model + review - —
default→Small model
Last 7 days · 412k requests
- Small model62%
- Large model31%
- Small + review7%
Blended cost $0.004 per request, was $0.007 with one model
What it does
Match each workload to the model that fits.
Use routing rules to make AI model choice a deliberate cost decision, not a fixed default.
- Send each AI request to the right model or endpoint using rules you set.
- Consider economic and workload requirements when defining the routing strategy.
- Review request distribution and cost effects using the telemetry available to each integration.
Data in
Cloud & operational data
Approved sources and scope
Analysis
Context & decisions
Rules, economics, ownership
Result
Measured outcomes
Review, validate, report
Why it matters
Define the requirements. Choose models against them.
Send each AI request to the right model based on cost, quality and uptime, and track the results over time.
Example
Compare models for a high-volume sorting task where an expensive model isn’t always needed.
Data and results
Routing decisions you can trace and explain.
Required data and dependencies
Integrated applications, available model endpoints, credentials and configured routing policies.
Operational outputs
Model-routing configuration, policy execution context and usage visibility.
Related capabilities
AI Token Caching & Reuse
Avoid unnecessary repeat inference costs
AI & LLM Cost Intelligence
Attribute the economics of inference
AI Spend Governance
Establish accountability for AI model consumption
See Policy-Based LLM Model Routing in context.
Check how this feature fits your cloud, data, permissions and business needs.
Availability note
The CloudScore product team confirms what’s available; setup depth and requirements depend on your deployment.
Book a demo
Try policy routing on your own traffic.
We’ll show this feature with examples like yours and explain the data and permissions it needs.