OpenAI says its models are now better at building and migrating software written in old programming languages, including COBOL, Fortran and Visual Basic. The company notes that 40 percent of the world's banks still rely on COBOL systems. Migrating to newer languages could cost them millions to hundreds of millions of dollars and take a long time.
The pressure is growing because the programmers who work with COBOL are aging out of the workforce. Banks still using the language will need to migrate within the next few years. OpenAI said improvements to its COBOL capabilities are guided by feedback from developers and enterprise customers.
To measure progress, OpenAI created an evaluation suite called Multi Lang Bench, which asks models to build SQLite-compatible database engines in a specified programming language. The company said GPT-6 Astra had an average test pass rate of 87.4 percent across 20 languages, compared with 26.8 percent for GPT-5.6 Sol. COBOL remained the hardest language, with Astra's pass rate at 72 percent.
OpenAI also tested whether models could migrate COBOL programs to C++, Java, Kotlin, Python or Rust while preserving their behavior. Accuracy matters because banking software must handle details such as rounding numbers with excess decimal digits and handling missing records. GPT-6.1 Sol achieved the highest accuracy at 97.9 percent with high reasoning effort, and 86.6 percent at low reasoning effort. OpenAI acknowledged that a migration succeeds only if "all tests pass," and said applying the technology to production software still requires expert review and testing against the system's business rules and dependencies.
Source: Engadget · Summarized by HeadlinesBriefing