OpenAI says new models improve COBOL, Fortran work

OpenAI says its new models perform better on legacy programming languages such as COBOL and Fortran. The company also released MultiLangBench on GitHub, an evaluation that tests whether models can build a SQLite-compatible database engine across 20 programming languages.

• OpenAI's new models improve work with COBOL and Fortran
• MultiLangBench evaluation released on GitHub
• Test checks building a SQLite-compatible engine in 20 languages