Xyntherium

Model Scorecards

A permanent record of how each model performs on the panel — how often it answers, how fast, how closely it tracks the verified consensus, and where it’s strongest. Every number is earned on live verification traffic.

Last updated: Jul 28, 2026

Scorecards are rebuilt daily from live verification traffic. The same per-task scores shown here are what the router weighs when it has more capable models than a panel needs — so the strongest model for a task is the one most likely to answer it.