https://derekwang2002.github.io/ai-intelligence/en/projects/openai-multilangbench/ Material changes in AI models, agents, engineering and infrastructure, with primary sources and recommendations. en Mon, 02 Nov 2026 16:56:00 GMT https://derekwang2002.github.io/ai-intelligence/en/events/ev-20261020-03/ ev-20261020-03-published Tue, 20 Oct 2026 16:04:41 GMT New TRIAL <p>OpenAI has published MultiLangBench, an Apache-2.0 benchmark covering 20 programming languages with less public training data, including COBOL, Fortran and Visual Basic. OpenAI reports GPT-6 Astra averaging an 87.4% test pass rate versus 26.8% for GPT-5.6 Sol. The repository contains public tasks and tests, but the headline model results are vendor-run and do not by themselves show production migration quality.</p><p><strong>Why read: </strong>Public tests add comparable baselines for languages such as COBOL and Fortran, without turning benchmark pass rate into migration success.</p><p>Recommendation: TRIAL</p><p>Sources: <a href="https://openai.com/index/low-resource-programming-languages/">Advancing coding in low-resource programming languages</a> · <a href="https://github.com/openai/MultiLangBench">OpenAI MultiLangBench repository</a></p>