Technology & AI

OpenAI says its models improved at migrating old COBOL code

OpenAI reported major coding gains across 20 programming languages, including a 97.9% accuracy score on a COBOL migration test. Banks and government agencies still rely heavily on such legacy systems.

By Standard Topic. AI-generated from the reporting listed under Coverage. Published . Updated .

This story is no longer being updated. It reflects reporting up to .

What happened

OpenAI said its newest models improved sharply at writing and migrating code, including in COBOL, an older programming language. The company reported a 97.9% accuracy score on a COBOL test for its GPT-6.1 Sol model, with gains across 20 programming languages. It also said it is adding evaluations for low-resource languages and legacy software tasks.

Why it matters

Banks, government agencies and other critical institutions run largely on antiquated programming languages, so better migration tools could help them update those systems.

Timeline

  1. Oct 19, 6:44 PM ET: OpenAI said it was improving ChatGPT for low-resource programming languages and adding evaluations for software engineering, SQLite and COBOL migration.
  2. Oct 20, 4:52 AM ET: OpenAI said GPT-6.1 Sol scored 97.9% on its COBOL test, with its newest models improving at writing and migrating code.
  3. Oct 20, 8:01 AM ET: OpenAI said its models now come with better COBOL capabilities that could help banks migrate away from old COBOL programs.

Coverage

More from Technology & AI

All Technology & AI news · Technology & AI story archive