24 August 2026

Nvidia's coding agent scores perfect on ARC-AGI-3 benchmark

First reported

AI Breakfast ran this on .

  • Nvidia's AVO agent completed all 183 levels of the ARC-AGI-3 benchmark without being given instructions, rules, or goals.
  • The agent inferred what it needed to do and adapted to unfamiliar tasks on its own, without explicit guidance.
  • This suggests the agent can understand objectives from context alone rather than relying on direct human instruction.

How it was covered

AI BreakfastIndependent editors

Nvidia's AVO coding agent scored 100% on the ARC-AGI-3 benchmark, completing all 183 levels without instructions, explicit rules or stated goals, suggesting the agent can infer objectives and adapt to unfamiliar interactive tasks.