AI Observatory, Runtime Observation Reports

Kimi K3 Under Observation – What 1,200 Executions Actually Reveal

NeoMundi Observatory ran a full runtime protocol on Kimi K3 (Moonshot): 1,200 valid executions – 120 situations repeated 10 times across 12 families (factuality, reasoning, governance, medical, legal, cybersecurity, business operations, and more). This is not a ranking.This is not a judgment.This is a runtime observation. The objective is to make visible what happens at […]