OpenAI reports big coding gains in COBOL and other rare languages

Summary

OpenAI says its latest models are far better at building and migrating software in low-resource languages such as COBOL, Fortran, and Visual Basic. On its own MultiLangBench evaluation, which asks for a SQLite-compatible engine in 20 languages, GPT-6 Astra passed 87.4% of tests on average versus 26.8% for the previous generation, and on a COBOL migration benchmark GPT-6.1 Sol reached up to 97.9% accuracy. All figures are OpenAI's self-evaluations, the evaluation suite is open-sourced on GitHub, and the company notes that production migrations still require expert review and testing.

Why it matters for our work

Critical systems in banking and public services still run on these older languages, so faster AI-assisted maintenance and migration could ease a long-standing bottleneck. Humans stay in the review loop while early-stage work speeds up.

Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.