OpenAI Says New Model Greatly Improves COBOL Code Migration
Understand the story
According to a report by The Verge on October 26, 2026, OpenAI said on October 19 that its latest model shows significant improvements in migrating legacy code written in older programming languages such as COBOL. The report cites OpenAI's test data: the new model GPT-6.1 Sol achieved a behavior preservation accuracy of 98% at high reasoning settings and 87% at low settings, compared with just 21% for GPT-6 Sol and 5% for GPT-5.6 Sol. The report also notes that experts have questioned the data and the claim, but did not disclose the specifics of the doubts. No independent verification from parties outside OpenAI has yet been seen, and the story is still developing.
AI overview of reporting · Updated 1 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- The Verge · AICan AI help solve the COBOL code crisis?
On October 19, OpenAI said its latest models showed significant improvements in migrating legacy-language code; in tests, GPT-6.1 Sol achieved 98% behavior-preservation accuracy at high reasoning effort and 87% at low effort, while GPT-6 Sol scored only 21% and GPT-5.6 Sol just 5%.
Story heat trend
Not enough continuous observations to show a trend yet.