TEST >> DECODE >> OPTIMISE >> COMPOUND
Live · scorecard - · refreshes 15m
The Smart Router Brain
Every model across OpenAI, Anthropic, xAI and Moonshot, scored live on accuracy, quality, speed and cost. For each task the brain chooses in order: the legal lane first - PHI, security, compliance - then the most efficient model within it .
CYORA Economics
Pure API
All Lanes
PHI-Legal
Offshore
● Free (on-shore)
● Flat (Max subs)
● Metered (API)
Efficiency = 100 × ( 0.30·Accuracy + 0.25·Quality + 0.15·Speed + 0.30·Cost ) each axis normalised 0-1 · the router re-weights the same four axes by task profile
CPM = blended USD per million tokens (3:1 in:out). Hosted models show API list price; on-shore models show an amortised self-host cost (GPU + energy), never zero - owning a model is cheaper, not free. Accuracy / Quality / Speed are reputational bands until the platform accumulates its own measured runs, then they auto-correct from the provenance ledger. The Efficiency score's cost axis reflects strategy ("CYORA Economics" credits paid-for Max subscriptions; "Pure API" costs everything at raw rate); the CPM column above is always the raw dollar truth. Data refreshes every 30 seconds from the 24/7 scorecard sentinel. Reports = data only, no PHI.