GPT-5.6 achieves only 7.8% on ARC-AGI-3 benchmark

OpenAI's GPT-5.6 Sol scored just 7.8% on the ARC-AGI-3 benchmark, while humans reliably score above 90%. The result highlights the gap between AI and human fluid intelligence, despite GPT-5.6's strong performance on other tasks like agentic coding and solving Erdős conjectures.