OpenAI’s COBOL migration score has made a striking leap
OpenAI released benchmark results showing its latest models have substantially improved at migrating COBOL programs and working with underdocumented legacy languages. GPT-6.1 Sol achieved 86.6% accuracy on the COBOL migration test at low reasoning effort—up from GPT-5.6 Sol's 5.4%—while reducing costs by 39%. At higher reasoning effort, the model reached 97.9%. OpenAI also released the MultiLangBench benchmark on GitHub, testing model capability to build SQLite-compatible database engines across 20 languages. However, the detailed evaluation of COBOL tests has not been publicly released, limiting independent verification. These results could help teams modernize and maintain legacy COBOL sys
Key points
- GPT-6.1 Sol achieved 86.6% accuracy on COBOL migration, up from 5.4% in prior model, with 39% cost reduction
- At higher reasoning effort, accuracy reached 97.9%
- OpenAI released MultiLangBench benchmark for testing model capabilities across 20 programming languages
- COBOL test details remain unpublished, limiting independent verification
Summary and key points written by AI from the headline and a short excerpt, so they can be incomplete or wrong. The New Stack did the reporting; read the original for the full story.
Read the full story at The New Stack
