OpenAI says its new models dramatically boost performance on low‑resource legacy languages such as COBOL, Fortran and Visual Basic
openai
| Source: Techmeme | Original article
OpenAI says its new models deliver major performance improvements for low‑resource legacy languages such as COBOL, Fortran and Visual Basic, which are used by banks and governments.
OpenAI announced that its latest model family delivers “major performance leaps” when working with low‑resource legacy languages such as COBOL, Fortran and Visual Basic – the codebases that still power many banks, government agencies and other critical infrastructure. The company says the new models are tuned to generate, debug and refactor code in these environments more reliably, and that they have added dedicated evaluations for software‑engineering tasks, SQLite handling and COBOL migration.
The claim matters because modernising legacy systems has long been a costly, high‑risk undertaking. Enterprises often rely on dwindling pools of specialists to maintain and upgrade decades‑old applications, and any tool that can automate parts of that work could shave months off migration projects and reduce the chance of costly errors. By extending its coding capabilities beyond the well‑served languages of today’s startups, OpenAI is positioning its platform as a practical assistant for the “brownfield” software that underpins finance, public services and regulated industries.
OpenAI’s rollout coincides with the broader launch of its GPT‑5.6 family, which includes three variants, and the controversial Astra model, underscoring a rapid expansion of heavyweight AI offerings. The next steps to watch are how quickly enterprise customers adopt the new capabilities, whether independent benchmarks confirm the touted gains, and how competing AI providers respond with their own legacy‑code tools. Regulators may also take note, given the sensitivity of the sectors involved and the potential for AI‑driven code changes to affect system stability and compliance.
Sources
Back to AIPULSEN