OpenAI has published COBOL and Fortran coding evaluations. GPT-6 Astra averaged 87.4% on MultiLangBench, a 20-language SQLite rebuild test. COBOL rose from 10.9% to 72.1% against GPT-5.6 Sol, and production use still needs expert review. https://www.developer-tech.com/news/openai-reports-87-4-multilangbench-pass-rate-gpt-6-astra/ #AI #SoftwareDevelopment #COBOL #LegacySystems