§ models · storyline
GPT-6's 99.9% ARC AGI score forces exam redesign for invention testing
GPT-6 scores 99.9% on ARC-AGI, prompting organizers to redesign the benchmark to evaluate invention-level capabilities.
GPT-6 Scores 99.9%: ARC AGI Exam Forced Remake, Next Level to Test "Invention" Capabilities 36Kr
§ sources1 publication · timeline below