GPT-6 Achieves 87% Success in Modernizing Legacy Code, Marking a Breakthrough by OpenAI

OpenAI Advances in Translating Legacy Code to Modern Languages

According to НВ — Техно: OpenAI has announced a significant breakthrough in converting software written in outdated programming languages into modern ones, a development crucial for updating systems used in banking, government agencies, and critical infrastructure. The organization developed two evaluation tests that showcased impressive results, especially with their GPT-6 and GPT-6.1 Sol models.

The first test, called MultiLangBench, measures AI’s ability to write SQLite programs across various languages. GPT-6 demonstrated an impressive success rate of about 87% on this benchmark.

The second assessment focuses specifically on migrating code from COBOL to modern languages such as C++, Java, Kotlin, Python, and Rust, while preserving the original program’s behavior. GPT-6.1 Sol nearly achieved a flawless performance, with results close to 100%. COBOL, developed in the late 1950s, remains widely used today in banking operations, including ATM systems. Meanwhile, Fortran continues to play a vital role in nuclear reactor simulations.

Market Implications and the Future of Software Development

As veteran programmers skilled in legacy languages retire, newer developers increasingly rely on AI tools capable of handling contemporary languages. Anthropic, another AI company, is also working on Claude Code, a tool aimed at modernizing COBOL-based software. According to Anthropic’s estimates, AI-driven modernization could shorten projects that once took years down to just a few quarters.

This technological progress had a notable impact on financial markets: in February 2026, following news about advancements in software modernization, IBM’s stock plummeted 13%, marking its largest single-day drop since 2000. This reflects how emerging technologies are disrupting established market players.

“Advances in AI programming technology have the potential to revolutionize how legacy software is maintained, which is critical for supporting essential infrastructure,”

Updating systems still reliant on COBOL and other legacy languages promises substantial reductions in both costs and time required for upgrades. Beyond boosting efficiency for financial institutions, this shift also lowers security risks associated with outdated technology.

As advancements in AI continue to reshape the landscape of software development, the accuracy of code migration tools is becoming increasingly critical. A recent report highlights how OpenAI’s GPT-6 has achieved remarkable accuracy in transitioning legacy programming languages, indicating a significant step forward in modernizing outdated systems. This progress not only benefits developers but also impacts industries reliant on legacy code, paving the way for more efficient and sustainable software solutions.