OpenAI publishes LifeVal, a benchmark of everyday tasks
On 2 November 2026 OpenAI published LifeVal. The page shows workflow examples, not a benchmark average. The new survey had 3,995 respondents.
OpenAI published LifeVal on 2 November 2026. It is a benchmark of everyday tasks where people said they wanted help from an assistant, not a new model. The page says the company used this evaluation over the past year to measure and push model performance on those tasks.
The build starts from the American Time Use Survey and the Multinational Time Use Study, plus a new survey of 3,995 respondents. Named areas include caregiving, health, education, household administration, personal finances, shopping, and government services. The page says more than one billion people use ChatGPT each week. That is OpenAI’s sentence.
The page shows workflow examples, and writes that the latest models score very highly on many LifeVal tasks. It does not publish a benchmark average in the text read. The examples on the page: kindergarten enrollment, 93.1%. Volunteers ahead of a flood, 98.0%. Home search, 100%. Watching a game with grandma from far away, 96.3%. Meal prep for a parent with a new diabetes diagnosis, 90.4%.
The page ties the high scores to training on browser use, computer use, and connectors, and says that pushes ChatGPT beyond chat. It does not say how many tasks LifeVal has, or which model produced each percentage.