OpenAI touts 87% COBOL migration accuracy with GPT-6.1 Sol

COBOL, created in 1959, remains deeply embedded in critical infrastructure. According to IBM, it is the foundation for more than 40 percent of online banking systems, supports 80 percent of in-person credit card transactions, and handles 95 percent of ATM transactions. One rough estimate puts the annual cost of COBOL's inefficiencies to the US GDP at around $105 billion.

OpenAI's blog post reported that GPT-6.1 Sol achieved 87 percent behavior preservation accuracy at low reasoning effort, climbing to 98 percent at higher reasoning, with comparisons to other models like GPT-6 Astra, GPT-6 Sol, and GPT-5.6 Sol. OpenAI spokesperson Oscar Haines acknowledged that legacy code is still far from a solved problem.

Experts who spoke to The Verge cautioned that such benchmarks are a long way from safe replacement of critical systems. Vadim Zaytsev noted that the largest system he'd seen had 250 million lines of COBOL, and that AI is not playing a major role at the moment in replacing or migrating legacy COBOL systems. Arie van Deursen described OpenAI's blog as superficial and stressed the gap between maintaining and modernizing. Alastair Donaldson highlighted that the main challenge is not understanding how the system works, and that nobody in their right mind would use COBOL for a new development project.