Model routing for agent cost control
Use cheaper models for routine steps and reserve expensive models for ambiguity, review, and synthesis.
Most agent workflows contain a mix of cheap and expensive reasoning. Model routing keeps premium models focused on the steps where they change the outcome.
AstraRunner treats this as an execution-platform concern: the run needs durable state, traceable decisions, cost visibility, and enough structure for teams to review what happened after the agent finishes.
Patterns that work
- Use small models for classification and formatting.
- Use stronger models for planning, conflict resolution, and final review.
- Escalate based on confidence and risk.
- Record routing decisions in the trace.
Implementation checklist
- Define model policy by role agent.
- Measure grade change by model tier.
- Cap expensive model retries.
- Fallback cleanly when a model is unavailable.
Watch outs
- Cheap models can create expensive downstream repairs.
- Routing policy needs eval coverage.
- Model names in prompts make migrations harder.
Related reading
How this fits the Astra stack
AstraRunner owns the operational path for agent work: orchestration, scheduling, approval gates, traces, costs, and role-agent handoffs. AstraMemory owns durable context and retrieval. AstraGenie turns those capabilities into product workflows for teams that want automation without rebuilding the platform layer.