Anthropic's new AI model can write code for 30 hours straight


Artificial intelligence software is not only behind chatbots, but also takes on programming tasks. A rival of the ChatGPT inventor OpenAI is now celebrating a major success. — Photo: Andrej Sokolow/dpa

SAN FRANCISCO: Anthropic – an AI company competing with the likes of ChatGPT maker OpenAI – has reached a milestone in using its software for programming.

According to the company, its new model, Claude Sonnet 4.5, was able to program software autonomously for around 30 hours straight. By comparison, the previous version of the software from earlier this year could focus on such tasks for about seven hours.

Rivalry in the fast-growing AI sector is not only about chatbot development, but also centres on which company can provide the most useful tools for software development.

AI tools have widely been adopted by programmers in efforts to speed up production, and in May ChatGPT maker OpenAI released a dedicated software engineering agent called Codex for software developers.

"Claude Sonnet 4.5 resets our expectations – it handles 30+ hours of autonomous coding, freeing our engineers to tackle months of complex architectural work in dramatically less time while maintaining coherence across massive codebases," said Anthropic chief executive Sean Ward.

According to Anthropic, Claude Sonnet 4.5 is particularly adept at identifying vulnerabilities in software code that could be exploited in cyberattacks. The company has already benefited from this capability in the development of its own software. – dpa

Follow us on our official WhatsApp channel for breaking news alerts and key updates!

Next In Tech News

How a cat named KitKat became San Francisco's latest symbol of anti-tech rage
Using AI instead of search means superficial learning, research shows
Microsoft now lets you save and synchronise passkeys in Edge browser
Opinion: What should you do if you lose your phone?
MCMC to summon Meta over reported profits from fraud, gambling ads
Inside the race to train AI robots how to act human in the real world
How a cat named KitKat became San Francisco's latest symbol of anti-tech rage
Stuck in traffic? Google Maps has a smarter AI assistant for you
Mistrial declared for MIT-educated brothers accused of $25 million cryptocurrency heist
Apple now letting users scale back its controversial new iOS design

Others Also Read