OpenAI releases a benchmark for less-common programming languages
MultiLangBench tests AI-built databases across programming languages; the research also examines legacy-code migration.
Announced October 19, 2026; the public GitHub test suite is readable, without an exact release time. This is research, not a new web or app feature. Vendor-reported gains do not establish production readiness; expert review and business-rule testing remain necessary. Documentation is English; no Traditional Chinese edition was announced. The repository uses Apache-2.0. OpenAI API access depends on account and region, with separate usage charges; no new pricing was announced. We have not tested it.
Usage conditions
- Hong Kong
- Public research and code do not establish API access in Hong Kong. It is absent from OpenAI's API support list. Local network access was not tested.
- Macau
- Public research and code do not establish API access in Macau. It is absent from OpenAI's API support list. Local network access was not tested.
- Taiwan
- Taiwan is listed for OpenAI API access, subject to account eligibility, model permissions and usage charges. Reading research is separate from running evaluations. Not tested.
Official source: OpenAI: Advancing coding in low-resource programming languages