OpenAI releases coding benchmark and reports sharp gains in COBOL

The company’s two evaluations separate building new software from translating old programs without changing their behavior. High migration scores still require expert review before production use.

Analyze with AIAnalyze with ChatGPTAnalyze with ClaudeAnalyze with PerplexityAnalyze with Google AI ModeAnalyze with GrokAnalyze with Microsoft Copilot