Western AIOpenAI Puts Low-Resource Code to a Byte-for-Byte Test
OpenAI's October 19 publication introduces two rigorous evaluations for coding in low-resource languages — MultiLangBench across 20 languages and a COBOL migration test demanding byte-for-byte output parity — revealing a dramatic performance gap between GPT-6 Astra and its predecessors, with implications for the trillion-dollar legacy modernisation market.
Lukas Hoffmann











