OpenAI says its latest models are much better at writing and migrating code in low-resource programming languages such as COBOL, Fortran and Visual Basic. These languages have far less public code and documentation for models to learn from than Python or JavaScript, yet still run critical systems.
On MultiLangBench, a new OpenAI test that asks a model to build a SQLite-compatible database engine from scratch in one of 20 languages, GPT-6 Astra passed 87.4% of tests on average at maximum reasoning effort, up from 26.8% for GPT-5.6 Sol. COBOL rose from 10.9% to 72.1%, Fortran from 21.4% to 83.5% and Visual Basic from 27.5% to 89.4%. COBOL is still the hardest of the 20.
The cheaper GPT-6 Luna reached a 40.3% average at an estimated API cost about 78% lower than GPT-5.6 Sol.
A second test asks models to migrate COBOL programs to C++, Java, Kotlin, Python or Rust, and counts a migration as successful only if hidden tests show identical outputs. GPT-6.1 Sol scored 86.6% at low reasoning effort, against 5.4% for GPT-5.6 Sol, at 61% of the cost. At higher effort it reached 97.9%.
OpenAI notes that COBOL supports more than 40% of online banking systems, citing IBM, and says using these models on production software still needs expert review and testing against each system’s business rules. It is releasing the MultiLangBench setup and test suite on GitHub.
Updating the ageing code behind banks could get cheaper, but these are OpenAI’s own benchmarks and results until others reproduce them.
Read the post: Advancing coding in low-resource programming languages