OpenAI Launches a Coding Model for Older Programming Languages

OpenAI today announced GPT‑6 Astra, an AI coding model tailored for COBOL, Fotran, Visual Basic, and other older programming languages.

“These languages have less code and documentation available for models to learn from than languages like Python and JavaScript, yet support critical infrastructure and essential services around the world,” the OpenAI announcement explains. “This work is guided by feedback from developers and enterprise customers who have told us that coding models often struggle with these languages.”

OpenAI developed two evaluations, MultiLangBench and COBOL migration, to ensure that its models can build new software in these older, less well-documented languages and preserve the behavior of older apps being modernized or otherwise migrated to newer languages. GPT-6 Astra scores 87.4 percent across 20 languages in MultiLangBench, up from 26.8 percent in GPT-5.6 Sol. And it scores 86.8 percent in COBOL migration–97.9 percent at high reasoning effort–up from just 5.4 percent for GPT-5.6 Sol.

“Our goal is to help models apply software engineering skills across more languages: Understanding unfamiliar repositories, planning changes, using build tools, running tests, and debugging failures,” OpenAI says.

GPT-6 Astra is available to developers through the OpenAI API on coding tasks for COBOL, Fortran, OCam, and Visual Basic.