A new agent framework reaches 85% on ARC-AGI3 public games

JFPuget · x · 2026-07-27

A new agent framework is being shared by the author’s colleagues, and the team says it reaches 85% on ARC-AGI3 public games.

The post suggests the framework includes a world model and memory component, based on the accompanying benchmark chart. The main signal is that the team is claiming a strong public-game result for an agent system, which makes this relevant both as an engineering update and as an eval data point.

Related event: New Agent Framework Scores 85% on ARC-AGI3 Public Games(2 posts)→

Original post →

More from coding & agent

coding & agent channel →