OpenAI Released GPT-6 Astra with Computer-Use Skills
The new AI model can navigate desktop interfaces to automate tasks on older software systems.
Updated on Sept. 22, 2026 in Artificial Intelligence

Live Poll
Do you trust artificial intelligence to manage sensitive systems without human supervision?
OpenAI has officially launched its GPT-6 Astra model, which features advanced computer-use capabilities by reading screens and interacting with programs. The technology is designed to assist industries by performing actions like typing and clicking within software interfaces.
Why it matters
Financial institutions are increasingly testing these AI agents as a workaround to avoid the high costs and operational risks associated with replacing outdated legacy banking infrastructure. This allows banks to modernize their workflows without fully rebuilding core systems.
GPT-6 Astra achieved a 72.6% score on the OSWorld 2.0 benchmark, completing tasks in an average of 40 minutes compared to 75 minutes for its predecessor. The model utilizes visual input to navigate interfaces, bypassing the need for fixed programming scripts.
The players
OpenAI
An artificial intelligence research organization that develops large language models and advanced automation software.
Goldman Sachs
A leading global investment banking and financial services firm currently evaluating AI for operational automation.
TSB Bank
A United Kingdom-based retail bank that faced significant regulatory fines following a troubled IT migration.
The details
The model functions by mimicking human computer interactions, including filling out digital forms and navigating through applications in real time. Banks are targeting these agents to manage systems built on COBOL, a programming language introduced in 1959 that still powers a significant portion of global financial infrastructure.
Timeline
1959: COBOL programming language was introduced.
April 2018: TSB Bank transitioned to a new platform.
February 2026: Goldman Sachs CIO conducted testing of AI agents.
June 2026: Benchmarking of AI agent performance was conducted.
September 2026: OpenAI released the GPT-6 Astra model.
The Tech Race
The introduction of agents capable of computer use marks a departure from traditional software integration by allowing AI to operate on top of legacy systems. This development challenges the necessity of full-scale infrastructure migration, which has historically been a slow and hazardous process for financial institutions.
For users and employees, this technology could automate repetitive data entry and navigation tasks within older workplace software suites. It may reduce the time required to complete administrative form filling while decreasing human errors during complex software interactions.
The takeaway
AI agents are transforming how corporations interact with legacy software by enabling automated navigation rather than costly infrastructure replacements. Organizations should prioritize assessing the security and reliability of these agents before allowing them to manage sensitive financial accounting processes.
Further reading
Learn more about the evolving capabilities of Artificial Intelligence.
Source note: This article includes information reported by PYMNTS.
Live Poll
Do you trust artificial intelligence to manage sensitive systems without human supervision?



