OpenAI models hit 97.9% accuracy in modernizing decades-old software
OpenAI reported that its coding models have made major gains in migrating legacy code in languages like COBOL and Fortran. GPT-6.1 Sol reached 97.9% accuracy on the company's COBOL migration test at high reasoning effort, versus 5.4% for GPT-5.6 Sol, while low-effort runs hit 86.6% with a 39% drop in API costs.
