Nvidia says its general-purpose coding agent system, AVO, scored 100% on the ARC-AGI-3 interactive reasoning benchmark’s public set, according to a Techmeme item citing Terry Chen on the NVIDIA Technical Blog. The reported result covers all 25 environments in the public set and all 183 levels. The cluster is narrow but notable because both visible items describe the result as involving the ARC-AGI-3 interactive reasoning benchmark. Hacker News also surfaced the Nvidia claim under the headline that AVO scored 100% on the ARC-AGI-3 interactive reasoning benchmark. The evidence available here does not include an independent benchmark audit or a third-party reproduction. Both visible items appear to point back to Nvidia’s own account of the result, so the correct reading is that Nvidia is reporting the score, not that outside evaluators have independently confirmed it. Techmeme’s feed summary includes a truncated excerpt saying the research project “elevates Claude Opus 5 from a 30% model baseline to,” but the provided material does not complete that comparison. Who benefits: Nvidia benefits from attention around a high-profile benchmark claim for AVO. Who's exposed: Too early to tell from the provided material.