model-routing TEC-0085Route routine calls to a smaller model
Send the short, structured, routine calls in an agent pipeline to a small cheap model and reserve a frontier model for the steps that need deep planning or sustained constraint tracking.
Production agent systems make many model calls per user request, and most are short and routine: format a payload, pick a tool, extract a field, check a condition. Routing sends those to a small open-weight model and escalates only the narrow class of steps that need frontier capability. The technique lives or dies on the router: it has to decide, before the work is done, which class a step falls into, and a wrong escalation decision costs either money or the task. Distinct from a quorum or ensemble, where several models answer the same question (multi-model-coordination), and from a capability budget, which bounds what an agent may do rather than which model does it.
Requires: At least two models of different cost, and a way to classify a step before running it. Cost savings assume the router is cheap relative to the frontier call it avoids.
Contexts: Autonomous agent, Coding agent
Does it work?
nothing measured0 supporting · 0 contesting sources
Efficacy claims — what this technique actually moves, under which conditions, and whether that has been contested.
1 more condition from claims reviewed by AI, listed with the claims below.
No efficacy claim filed yet. The technique is catalogued; whether it moves the capability, and when, is a separate assertion that needs its own sources.
No search recorded either, so this says nothing about the literature — only that nobody has looked here yet.
Code
No repository linked yet. Contribute one.
Sources
Whether a technique works is a claim. What you file here becomes a claim about this technique under the first capability it addresses, with your evidence as its source, and moves the standing above the same way a paper would.