Claude Opus 5 scores nearly four times higher than GPT-5.6 Sol on ARC-AGI-3

Claude Opus 5 has widened the gap on ARC-AGI-3, scoring 30.2 percent versus the previous record of 7.8 percent set by OpenAI’s GPT-5.6 Sol. It solved five environments that no earlier model had completed, with four reaching or exceeding reported human perform…


