November 3, 2026
OpenAI announced that its artificial intelligence technologies are showing increasing efficiency in performing daily human tasks, such as shopping, budgeting, and meal planning, according to a report published by CNN.
In a related context, the company unveiled a new internal tool that aims to evaluate the ability of its models to deal with real-life tasks, in conjunction with publishing detailed research on the level of performance of these models. OpenAI explained that its model, called “GPT-6 Astra,” recorded high accuracy rates, reaching 93.1% in searching for a suitable kindergarten after moving to a new area, and 98% in managing local volunteers to respond to a flood crisis, in addition to achieving 90.4% in designing a meal plan for a parent who was recently diagnosed with diabetes.
These developments come at a time when competition is intense between major technology companies, such as Meta, Anthropic, and Google, to develop artificial intelligence agents capable of carrying out practical tasks in the real world. At the same time, OpenAI and its competitors continue to face fundamental challenges related to the safety of this technology, in addition to financial pressures to achieve profits that cover the volume of huge investments in this sector. Before revealing the latest evaluation results, the company sought to find accurate measurement mechanisms to determine the efficiency of “GPT Chat” in accomplishing practical tasks, such as preparing purchase lists and filling out online forms.
In statements to CNN before announcing the results of the evaluation, Jenny Yang, a researcher at OpenAI, said: “Given that a large number of people use the application, we seek to monitor various use cases and the ways in which users employ this application, to understand precisely how we achieve improvement over time, what aspects we still face difficulties in, and how we can develop this technology to become more useful to people.”
Standardized assessments are considered an essential element in the artificial intelligence development sector, as laboratories rely on them as a reference to train models and improve their performance in specific areas. It is noteworthy that last year, OpenAI launched a standard known as “GDPVal” to measure the performance of artificial intelligence in skills related to work environments across various sectors, including manufacturing, government work, finance, and sales. In terms of expanding infrastructure, the company plans to purchase chips from Google to enhance its computing capabilities.
