Today· 22 JulRelease

Adaptive Inference Launch

Medium5.5

◆ Description

Marathon becomes available as adaptive inference infrastructure for long-running agents. Each request supports a completion window from now to anytime, with up to ~65% lower cost when users can wait.

◆ Why it matters

The compute layer may broaden KITE AI's utility by making agent inference more programmable and affordable. Investor relevance is limited until usage or a direct token connection is demonstrated.