SmartGPT: Major Benchmark Broken - 89.0% on MMLU + Exam's Many Errors | AI Explained | Podwise