OpenAI has initiated work focused on refining model performance for programming languages that have traditionally seen limited representation in training data. The initiative centers on "improving ChatGPT across low-resource programming languages" to support a wider range of development environments. Alongside these adjustments, the company is implementing "new evaluations for software engineering" to better measure coding efficacy.
The updated assessment framework targets specific technical domains that developers face in production and legacy maintenance. According to the update, these targeted testing benchmarks specifically cover "SQLite, and COBOL migration." These evaluations seek to measure how the model handles both specialized database operations and older enterprise codebases that require modernization.