OpenAI's GPT-6.1 Sol model achieves success in transforming legacy software languages
The GPT-6.1 Sol model, developed by OpenAI, has achieved high accuracy rates in the process of transferring legacy programming languages such as COBOL and Fortran into up-to-date languages. The model aims to contribute to the modernization of critical systems at a lower cost compared to previous versions.
OpenAI's new GPT-6.1 Sol model demonstrated significant performance in converting legacy programming languages such as COBOL, Fortran, and Visual Basic used in critical infrastructures like banking and ATMs into modern languages.
The decreasing number of experts knowing legacy software languages made it difficult to update systems written in these languages. While the GPT-6 Astra model showed 87.4 percent success in writing databases from scratch in 20 different languages in MultiLangBench tests, the GPT-6.1 Sol model achieved 86.6 percent accuracy in conversion tests of COBOL-based software without disrupting their operations. When the model's reasoning capacity was increased, this rate rose up to 97.9 percent.
GPT-6.1 Sol can process at a 39 percent lower cost compared to previous models. Experts state that despite the technical advancement provided by the model, human oversight and additional testing processes are mandatory in real-world applications.
Sources
- cnnturk.com · Link