OpenRouter launches Auto Router, Auto Code

OpenRouter is turning model selection into adaptive infrastructure, but the real test will be whether dynamic routing stays reliable across developers’ varied workloads.

• –Auto Code tracks edits, test failures, recurring errors, pass rates, speed, and cost to guide model choices across agent sessions.
• –The routers weigh eleven request-level factors alongside live pricing, provider health, cache status, and governance restrictions.
• –Routing insights expose why a model was selected, improving visibility into an otherwise opaque layer of AI infrastructure.
• –The 74% cost reduction and benchmark parity are OpenRouter-reported claims, so teams should validate them on representative repositories.
• –Dynamic routing trades some reproducibility and peak-model control for lower maintenance, better cost efficiency, and automatic adaptation as models change.