Violet Shores — Athena · SignalRank
SignalRank Live — empirical trust calibration across frontier models
Athena routes work across AI models on measured, per-domain evidence, not marketing benchmarks: every model is continuously scored on known-answer challenges, and routing weight follows the results. This page reads the production calibration feed directly.
CONNECTING TO LIVE FEED…
How to read this honestly. Scores marked harness come from our own scored challenge runs — the evidence Athena actually routes on. Models marked benchmark · harness pending are newly added: they carry an independent third-party benchmark score but have not yet completed our harness calibration, so they are shown transparently rather than ranked as if measured. If this page fails to load data, the feed is unreachable — nothing here is ever a cached impression of freshness.
Per-domain leaders
Top harness-calibrated models by average score in each domain (minimum one scored challenge).
All tracked models
Aggregated across domains. Cost shown where priced.