GoKawiilOpenAI touts improved COBOL coding and migration performance in new models
OpenAI released benchmark results showing its latest models, including GPT-6 Astra and GPT-6.1 Sol, perform significantly better at writing and migrating code in legacy languages like COBOL, Fortran and Visual Basic. Using an internal test suite called MultiLangBench, OpenAI reported Astra's average pass rate across 20 languages rose to 87.4 percent, though COBOL remained the toughest language at 72 percent. Separately, GPT-6.1 Sol reached up to 97.9 percent accuracy when migrating COBOL programs to languages such as Python, Java and Rust while preserving original program behavior.
GoKawiil's interpretation of the reporting above, not reported fact.
OpenAI frames this work around banks, noting that roughly 40 percent still run COBOL systems and face costly, time-consuming migrations as COBOL programmers retire. Improved model accuracy at preserving exact program behavior—like decimal rounding or handling missing records—could make AI-assisted migration more viable for risk-averse financial institutions, though OpenAI itself acknowledges success requires all tests to pass, a high bar for banking-grade reliability. The benchmarks are self-reported by OpenAI, so independent verification of real-world migration outcomes is still needed.
- OpenAI's GPT-6 Astra scored 87.4 percent on MultiLangBench across 20 languages, up sharply from GPT-5.6 Sol's 26.8 percent.
- COBOL remains the hardest language for AI models, with Astra scoring 72 percent versus higher scores elsewhere.
- GPT-6.1 Sol hit 97.9 percent accuracy migrating COBOL to modern languages while preserving program behavior, under high reasoning effort.
Source: engadget.com — Mariella Moon, 2026-10-20
Published there as: “OpenAI says its models can help banks migrate from old COBOL programs”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.