Frontier Model Cost and Open-Weights Popularity is Driving Demand for Model Routing

Model routing has become a crucial component of AI deployment due to intense competition among frontier model companies and the increasing power of open-weight models. Enterprises are heavily adopting this approach primarily for economic reasons, as advanced frontier models and longer tasks have caused per-user costs to skyrocket significantly over the past year.

Glean, an enterprise AI platform valued at $7.2 billion with $300 million in annual recurring revenue, operates as a meta-harness combining leading large language models into a single experience. The platform offers three levels of model selection: manual employee choice, administrator restrictions, and an automatic mode that dynamically selects a model for each task. This automatic routing, combined with its architecture, makes Glean significantly more cost-effective than competing tools.

A key part of Glean’s architecture is an agentic search model called Waldo, which gathers raw materials and filters user queries before handing off complex work to frontier models. This process reduces latency and saves tokens, reserving expensive models only for tasks that truly require them. Furthermore, because AI usage has grown expensive, enterprises have rapidly shifted toward incorporating open-weight models, which are an order of magnitude cheaper, into their core strategies.

Glean continuously refines its model routing through a human feedback loop and internal testing systems. By observing how ordinary business users interact with the platform and running background comparisons using AI-based judges, Glean evaluates router performance and trains the system on real-world traffic.