OpenAI has launched its GPT-6 models, showcasing significant improvements in handling legacy programming languages like COBOL and Fortran. According to The Decoder, the GPT-6 Astra model achieved an 87.4 percent test pass rate across 20 languages, a notable increase from the 38 percent pass rate seen in the GPT-5.6 Terra version.
Focusing specifically on COBOL migration tasks, the GPT-6.1 Sol model demonstrated up to 97.9 percent accuracy, underscoring the new model's enhanced capabilities in supporting older, yet critical, enterprise systems. These advancements could accelerate modernization efforts in industries still reliant on legacy codebases.
For Japanese markets, where financial institutions and government bodies continue to depend on legacy systems, OpenAI’s progress in legacy language processing may offer new tools for streamlining digital transformation and maintaining competitiveness in technology-driven sectors.