Abstract
Existing conversational artificial intelligence systems can face challenges in balancing responsiveness, computational capability, and operational cost. Architectures that rely on a single large model may introduce latency and high costs, while smaller local models may lack the capability for complex reasoning. This disclosure describes a multi-tiered model architecture that can dynamically route user requests. A routing mechanism can analyze an incoming query and direct it to an appropriate computational tier. For example, simple requests may be handled by a lightweight, on-device model on a computing device (e.g., a smartphone, smart watch, wearable device, etc.) for low-latency responses, while more complex or long-running tasks can be escalated to larger, more powerful models, which may be hosted remotely. This approach of selectively engaging different model tiers based on task complexity may provide a responsive user experience while managing computational resources and retaining the capability to handle sophisticated requests.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Pedersen, Zebedee; Yip, Beverly; Griffith, Jordan; and Tanimura, Hana, "Multi-Tier Model Architecture for Dynamic Request Routing in Conversational AI", Technical Disclosure Commons, (November 10, 2026)
https://www.tdcommons.org/dpubs_series/12949