---
title: "Wayfinder — pre-deployment testing for LLM inference"
description: "Wayfinder finds the best deployment configuration before you commit. Programmable testing, comparative analysis, and theoretical roofline reference, in one place."
url: "https://iop.systems/wayfinder/"
---

# Wayfinder

*Tune your stack before you deploy.*

Operating an inference stack involves an enormous number of decisions. Wayfinder helps navigate that space to find the most cost-effective deployment for your model, and shows you how close your stack is to the hardware's theoretical limits.

## The problem
The inference ecosystem evolves rapidly. Every updated model, runtime, or hardware release is a potential tripwire: a configuration that was optimal last month may be leaving performance — and money — on the table today.

## What Wayfinder does
- **Programmable testing** — describe the sweeps you care about across runtimes, hardware, and configurations, and run them reproducibly.
- **Comparative analysis** — put configurations side by side to see the real trade-offs in latency, throughput, and cost.
- **Roofline reference** — evaluate results against the hardware's theoretical performance ceiling, so you know how much headroom is actually left.

Wayfinder is part of Inference², from IOP Systems. See also [Sightlines](https://iop.systems/sightlines/) for production observability.