OpenAI launches LifeVal everyday-task benchmark
OpenAI has released LifeVal, an internal benchmark of everyday agentic tasks built from time-use surveys, showing how it measures ChatGPT's progress toward becoming a personal assistant that gets real tasks done.
OpenAI announced LifeVal, an evaluation built to measure how well its models handle practical, everyday tasks rather than academic knowledge tests. The company says it has used the benchmark over the past year to track and improve model performance.
To choose the tasks, OpenAI started with two large studies of how people spend their time: the American Time Use Survey and the Multinational Time Use Study. It then ran its own demographically representative survey of 3,995 respondents to find where people most wanted help from a personal assistant. The resulting tasks cover caregiving, health, education, household administration, personal finances, shopping and access to government services. Examples include finding childcare, making a budget, changing a recipe, ordering groceries and planning workouts.
OpenAI frames the eval as part of an effort to move ChatGPT "beyond a chat tool" toward an agentic assistant that takes actions. The company says it has spent the past year training its models to get faster and more capable at browser use, computer use and connectors tasks.
The announcement gives few details that would let outsiders check the results. OpenAI says only that its latest models "score very highly on many LifeVal tasks." It publishes no aggregate scores, comparisons with earlier OpenAI models or competitors, or methodology for grading. The post also does not say how many tasks the benchmark contains or whether the dataset will be released. For now, LifeVal is an eval that OpenAI built and reports on itself.