Claude Sonnet 4.5, a model that he can create programs and complete work tasks independently

The company said the model was able to operate independently for 30 hours, while maintaining a constant focus with minimal control while building a full software application. It is a great improvement in the company’s previous OPUS 4 model, which was released four months ago, which can work independently for only seven hours.

Anthropor said that Claude Sonnet 4.5 outperformed OPUS over the main standards and was more effective in meeting the practical work needs of customers. The company said the model was better in coding than previous border models, and that it is on the latest Swe-Bench, a major standard that tests how models are performed in software development tasks. Anthropor said that Claude Sonit 4.5 was better than its predecessors in the following instructions, identifying the improvements of code, and generating more software instructions ready for production. When tested at tasks from the financial services industry, the company said that the new model surpassed the forms of Claude in tasks such as research, building financial models and prediction.

Antarubor seems to pushing more before its competitors in helping coding and completing independent tasks, which puts his models towards the use of companies and in the workplace. The former Claude Obus 4.1 model has already defeated competitors New Openai Standard From completing the professional task, the gross domestic product, which tested how to perform models compared to human professionals through a group of industries and jobs.

Last week, Openai said that the GPT-5 and Claud Opus 4.1 model “are already approaching the quality of work produced by industry experts.”

Fencing Using studies It was also released earlier this month also that Claude models in humans have emerged models of the most professional artificial intelligence, especially compared to the Chatgpt from Openai, which are increasingly used as a consumer product.

According to the study, most Claude users resorted to models of the workplace or productivity tasks, where mathematical tasks and coding were martyred as dominant activities worldwide Claude.

The commercial use of Claude was highly based on assigning tasks. According to the study, approximately 77 % of the claims received by the model through its application programming interface – the application programming interface that is used primarily by institution agents – users demand to demand tasks on their behalf, instead of providing advice or suggestions. These reactions that focus on business are also concentrated in coding, which represent 44 % of API use. Another 5 % of API use is allocated to develop or evaluate artificial intelligence systems.

The tasks that business users are also automated tend to be the most expensive to run. The results indicate a shift in how companies deal with these tools. Instead of using it mainly to support decision or research, many differences depend on them to take the work completely from their paintings.

If models such as Claude are able to become more able to work independently, especially in the complex and intensive areas of time such as software engineering, the effects of companies and employees may be important. Independent factors can reduce the need for continuous human supervision and reduce costs on the repeated workflow, which speeds up the company’s operations and may reduce the need for the number of employees.

Fortune Global Forum Returns 26 to 27 October, 2025 in Rydah. Executive chiefs and world leaders will meet for a dynamic event for the call only forms the future of business. Apply for an invitation.