Glean Auto Routing picks the right AI model per task, backed by 5,000 enterprise evaluations

verrsane · x · 2026-11-12

Glean launched Auto Routing, which matches every request to the right model based on reasoning level, quality, modality, complexity, cost and task type — balancing speed, answer quality, and cost. The policy is backed by evaluations across 5,000 real enterprise tasks, with Cost-efficient / Balanced / Frontier modes and a model pool spanning Claude Opus 5, Claude Sonnet 5, Gemini 3.6 Flash, and GPT-5.6 Luna/Sol.

The poster calls it the best auto-routing policy he has seen, letting Glean pick the model 90% of the time so employees can just get work done.

Original post →

More from Apps

Apps channel →