← Back to feed

KradleBench – Multi-agent AI evals in Minecraft

AI · Studies · 2026-11-18 19:14

Kradle has launched KradleBench, a dynamic multi-agent AI evaluation benchmark set in Minecraft. In this new test, GPT-6.2 Astra scored 78.3 points, surpassing human expert SethBling to take the top spot.

Source: www.kradle.ai?1


Auto-generated summary for reference only. RSS