# Preferences Over Benchmarks: Model Routing — Archana Kamath & Tyler Gillam, DigitalOcean

Source: [AI Engineer](https://www.youtube.com/watch?v=FvxY8oPoI8o)  
Feed7 permalink: https://feed7.dev/p/preferences-over-benchmarks-model-routing-archana-kamath-tyler-gillam-di-06nxtgu  
Published: 2026-08-22T15:30:18.000Z  
Trust: Source Linked (source_linked)

## Why Included

Per-task model routing cut the demonstrated coding session’s cost from 44¢ to 14¢ with similar completion time, but builders still need workload-specific evals to validate quality.

## Source Summary

DigitalOcean’s demo routed code generation, testing, and documentation to configured models while a comparison session used Opus. The routed session cost **14 cents versus 44 cents**, and its routing decision was reported at **under 200 milliseconds**.

## Practical Implication

Define routing around task, latency, cost, preferred models, and hard failover rules. Evaluate the full coding workflow on your own cases, then adjust model pools and policies instead of choosing one model from a public leaderboard.

## Agent-Ready Context

DigitalOcean’s demo routed code generation, testing, and documentation to configured models while a comparison session used Opus. The routed session cost **14 cents versus 44 cents**, and its routing decision was reported at **under 200 milliseconds**.

Define routing around task, latency, cost, preferred models, and hard failover rules. Evaluate the full coding workflow on your own cases, then adjust model pools and policies instead of choosing one model from a public leaderboard.

The app comparison was a live demonstration and visual judgment, not a controlled quality study. A separate evaluation scored router correctness at 90% versus 95% for Opus, which the speakers described as close, but no judge uncertainty was quantified.

## Connected Context

Feed7 judgment across 545 accumulated Signals:

This adds a workflow-level demonstration that routing different coding stages can reduce observed cost and keep routing latency small, while confirming that model choice should follow local preferences and cases rather than leaderboards. It does not establish equivalent output quality: the comparison was visual, and the reported router scores lack judge-uncertainty analysis, so the result supports experimentation and policy tuning rather than a default route.

- [Notion's Token Town — Sarah Sachs, Notion](https://feed7.dev/p/notion-s-token-town-sarah-sachs-notion-1rb6gwh) — It reinforces Notion’s recommendation to route by task and preserve optionality, adding a concrete multi-stage coding demo while retaining the need for workload-specific economic evaluation.
- [TRACE-ROUTER: Task-Consistent and Adaptive Online Routing for Agentic AI](https://feed7.dev/p/2607-22465v1-1g7nw7j) — Both evaluate task-aware routing, but this demo assigns different coding stages to configured models whereas TRACE-Router keeps calls for an agent task on one backend and learns from final outcomes.
- [AI Gateway adds unified fast mode support](https://feed7.dev/p/ai-gateway-adds-unified-fast-mode-support-144dq26) — Unified fast mode is an implementation control for the target’s latency preferences and fallback rules, though it does not replace local workflow-quality evaluation.

## Context Map

- Layer: infra
- Domains: coding
- Topics: model-selection, gateways, agent-evals

## Uncertainty

- The app comparison was a live demonstration and visual judgment, not a controlled quality study. A separate evaluation scored router correctness at 90% versus 95% for Opus, which the speakers described as close, but no judge uncertainty was quantified.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.