OpenAI enhances model capabilities to build and migrate legacy code
OpenAI announced model upgrades designed to build and migrate low-resource legacy programming languages including COBOL, Fortran, and Visual Basic.
OpenAI announced that its latest model advancements feature significantly improved capabilities for building and migrating software written in low-resource programming languages, highlighting COBOL, Fortran, and Visual Basic. The company noted that these older languages continue to support mission-critical software used by banks, governments, and enterprises, making it necessary to accurately interpret decades of business logic when modernizing systems.[1][3][4][7]
To gauge performance, OpenAI introduced benchmarks that test software development from scratch across 20 languages and evaluate the translation of legacy code into modern languages such as C++, Java, and Rust, according to webulite. On the COBOL migration test, newer models reached up to 97% accuracy with increased reasoning effort, up from under 6% in prior versions. The upgrade targets infrastructure where COBOL still powers over 40% of online banking systems and 95% of ATM transactions, alongside Fortran applications in scientific modeling and weather forecasting.[6][5]
The improvements are aimed at reducing modernization expenses and helping organizations manage a shrinking talent pool of legacy engineers, Crypto Briefing reported. OpenAI said it designed the features in response to enterprise client needs and is asking engineering teams to submit requests for additional workflows and languages to target next.[1][2]
Key facts
- OpenAI announced improved model performance for building and migrating low-resource legacy languages, specifically COBOL, Fortran, and Visual Basic.
- OpenAI established benchmarks across 20 languages to evaluate greenfield coding and legacy translation into modern targets such as C++, Java, and Rust.
- Newer models scored up to 97% accuracy with higher reasoning effort on a COBOL migration benchmark, compared to below 6% for older models.
- COBOL powers more than 40% of online banking systems and 95% of ATM transactions, while Fortran remains essential for weather forecasting and scientific modeling.
- The advancements could lower modernization costs and mitigate shortages of engineers skilled in legacy software, according to Crypto Briefing.
- OpenAI stated it is tailoring these tools to enterprise feedback and invited developer teams to propose subsequent languages and workflows.
Sources · 7 sources
- OP
OpenAI@OpenAIPost on X ·
Our latest model advances bring stronger capabilities for building and migrating software in low-resource programming languages such as COBOL, Fortran, and Visual Basic. These languages power software that banks, governments, and businesses depend on. Progress here means understanding decades of business logic and helping developers modernize those systems while preserving what works. We’re building these capabilities around the challenges developers and enterprise customers bring us. If your team is working on these systems, tell us which languages and workflows we should tackle next. https://t.co/2TbnkQixJ1
Open source - CB
Crypto BriefingArticle ·
OpenAI upgrades its model for COBOL, Fortran, and other legacy code OpenAI's advancements in legacy code handling could significantly reduce modernization costs and address the shrinking pool of skilled engineers. The post OpenAI upgrades its model for COBOL, Fortran, and other legacy code appeared first on Crypto Briefing .
Open source - TE
Techmeme@TechmemePost on X ·
OpenAI says its new models show major performance leaps in low-resource legacy languages like COBOL, Fortran, and Visual Basic, used by banks and governments (OpenAI) (Visit Techmeme dot com for the link and full context!)
Open source - TE
TechmemeArticle ·
OpenAI says its new models show major performance leaps in low-resource legacy languages like COBOL, Fortran, and Visual Basic, used by banks and governments (OpenAI) OpenAI : OpenAI says its new models show major performance leaps in low-resource legacy languages like COBOL, Fortran, and Visual Basic, used by banks and governments — Our latest model advances bring stronger capabilities for building and migrating software in low-resource programming languages such as COBOL, Fortran, and Visual Basic.
Open source - GI
Gizmodo@GizmodoPost on X ·
OpenAI Says Its Models Can Give the Financial Sector a Coding Makeover https://t.co/b8nwoZpT6l
Open source - WE
webulite@webulitePost on X ·
OpenAI announced significant performance improvements in handling low-resource legacy programming languages like COBOL, Fortran, and Visual Basic. These older languages have far less training data and online documentation available than modern languages like Python, yet they continue to underpin critical infrastructure worldwide. COBOL alone powers over forty percent of online banking systems and ninety-five percent of ATM transactions, while Fortran remains essential for weather forecasting and scientific modeling. To evaluate these advancements, OpenAI developed new benchmarks testing a model's ability to build software from scratch across twenty languages and to translate legacy codebases into modern languages like C++, Java, and Rust. On the COBOL migration benchmark, newer models achieved up to ninety-seven percent accuracy with increased reasoning effort, representing a drastic leap over older models that scored under six percent. These advancements aim to assist enterprise and government organizations facing a shortage of specialized legacy developers. By improving automated translation, code reasoning, and debugging for decades-old systems, the updated models offer organizations a path to modernizing critical financial and institutional software without risking core operational behavior or data accuracy.
Open source - EN
Engadget@engadgetPost on X ·
OpenAI says its models now come with better COBOL capabilities. https://t.co/MmJ63iXEvk
Open source