BIG LAB UPDATES · 20 OCTOBER 2026
OpenAI says its newest models are much better at old business code
OpenAI says its latest models are far better at building and migrating software in older or niche languages such as COBOL, Fortran and Visual Basic, which still run a lot of banking and government systems. On its own tests, GPT-6.1 Sol moved old COBOL programs to a modern language with 86.6% accuracy at low effort and 97.9% at higher effort, up from 5.4% for an earlier model. These are OpenAI's own benchmarks.
Why it mattersIf your business depends on an old system nobody wants to touch, from an accounting package to a custom Visual Basic tool, AI-assisted migration is becoming realistic. Start with a small, well-tested piece and check its behaviour matches the old one before moving anything important.
Sources
How we knowCompany figures are labelled as their own claims; anything we couldn't verify is labelled as a report or a claim.
MORE FROM 20 OCTOBER 2026
- MiniMax M3.1 Flash: a very cheap model for running agents, with one big caveat
- The US orders DeepSeek out of federal agencies, including tools built on it
- Anthropic, Google and OpenAI are all reportedly lining up new models
- Meta explains how it trained its small open agent model, Muse Glimmer
- A free library of about 50 marketing skills for your AI agent
- Hy-Memory 2.0: one shared memory for all your AI coding tools