OpenAI says new models code better in COBOL and other low-resource languages
OpenAI says its latest models are much better at building and migrating software in low-resource languages such as COBOL, Fortran, and Visual Basic, which have less training code than Python and JavaScript. The company says COBOL supports more than 40% of online banking systems and 95% of ATM transactions, so better AI tools could help teams maintain systems that essential services depend on. On MultiLangBench, a test of building a SQLite-compatible database engine in 20 languages, OpenAI says GPT-6 Astra averaged an 87.4% pass rate, up from 26.8% for GPT-5.6 Sol. On a COBOL migration test, OpenAI says GPT-6.1 Sol reached 86.6% accuracy at low reasoning effort, compared with 5.4% for GPT-5.6 Sol, and OpenAI is releasing the MultiLangBench test suite on GitHub.