OpenAI Releases GPT-6 Astra with 97.9% COBOL Migration Accuracy
OpenAI released a research report stating that its enhanced model, GPT-6 Astra, achieves breakthroughs in building and migrating low-resource programming languages like COBOL and Fortran. On the newly introduced MultiLangBench, GPT-6 Astra reached 97.9% accuracy in migrating COBOL to modern languages, significantly outperforming previous models and aiming to support legacy system modernization.
Full Report
OpenAI Enhances AI Models to Tackle Niche Programming Languages, COBOL Migration Accuracy Soars to 97.9%
On October 21, it was reported that OpenAI released a research study yesterday (October 20), highlighting the enhanced capabilities of models like GPT-6 Astra in building and migrating "low-resource programming languages" such as COBOL, Fortran, and Visual Basic. Although these languages have limited public code and documentation, they underpin critical systems including banking, ATMs, weather forecasting, and nuclear safety analysis.
To measure these capabilities, OpenAI introduced two benchmarks: MultiLangBench requires models to build a SQLite-compatible database engine from scratch in 20 different languages; the other benchmark tasks models with migrating COBOL programs to languages like C++, Java, Python, and Rust while preserving original business logic.
The results show that GPT-6 Astra achieved an average test pass rate of 87.4% across the 20 languages in MultiLangBench, a significant improvement over GPT-5.6 Sol's 26.8%. Specifically, performance in COBOL rose from 10.9% to 72.1%, Fortran from 21.4% to 83.5%, and Visual Basic from 27.5% to 89.4%.
The more cost-effective GPT-6 Luna achieved an average pass rate of 40.3%, with estimated API costs approximately 78% lower than GPT-5.6 Sol. IT Home has attached relevant images below:
In the COBOL migration benchmark, GPT-6.1 Sol achieved an accuracy of 86.6% at low reasoning intensity, far exceeding GPT-5.6 Sol's 5.4%, while reducing costs by 39%. At higher reasoning intensity, accuracy reached 97.9%, approaching the evaluation ceiling.
OpenAI stated that this progress helps enterprises maintain legacy systems, fix defects, and advance software modernization. However, the company emphasized that using AI-driven migration results in production environments still requires expert review, combined with testing against specific business rules and dependencies.
