GLM 5.2 Matches Human Accuracy on VAT Bookkeeping Benchmark

GLM 5.2, a new AI model, achieved 98.7% accuracy on a VAT bookkeeping benchmark, nearly matching human performance. The model was tested on 10,000 real-world invoices and outperformed previous AI systems by 12%. This marks a significant step toward automating complex financial compliance tasks.