NVIDIA AVO agent scores 100% on ARC-AGI-3 benchmark

NVIDIAAI · x · 2026-08-21

NVIDIA's general-purpose coding agent, AVO, achieved a 100% score on the ARC-AGI-3 interactive reasoning benchmark. It completed all 183 levels across 25 public environments without instructions, explicit rules, or stated goals. The architecture integrates persistent memory, supervision, and tool use to elevate baseline model performance from 30% to 100%.

Related event: NVIDIA's AVO Agent Scores 100% on ARC-AGI-3(2 posts)→

Original post →

More from coding & agent

coding & agent channel →