# Ensemble Routing for Prompt-Adaptive Cost-Aware LLM Inference > Source: > Published: 2026-11-12 06:00:22+00:00 Large language models (LLMs) increasingly support Web-based services that must balance response quality and inference cost across heterogeneous requests. Conventional routing selects one model per prompt, while fixed ensembles cannot adapt their composition and cost to individual requests. We introd