The Pulse
OpenAI Launches LifeVal to Measure AI on Everyday Tasks
OpenAI introduced LifeVal on November 2, 2026, an evaluation of AI performance on practical tasks such as childcare, budgeting, shopping and accessing government services. The company says GPT-6 Astra earned 100% on several listed tasks, in
AI.info Team · Written with AI · fact-checked ·
From answers to getting things done
OpenAI introduced LifeVal on November 2, 2026, to measure how its models handle everyday tasks that can require more than producing a useful-sounding answer. The evaluation covers practical requests such as finding childcare, preparing for school, modifying recipes, ordering groceries, creating a budget and accessing government services. OpenAI frames the test around a personal assistant’s ability to carry out actions, not only offer advice.
“To be a truly helpful personal assistant, ChatGPT should be able to take actions that actually get things done,” the company writes. OpenAI says it used the evaluation over the past year to track and drive improvements on everyday tasks. The announcement places LifeVal alongside its push to expand ChatGPT’s browser use, computer use and connector capabilities.
A survey shaped the task set
OpenAI says it began with the American Time Use Survey and the Multinational Time Use Study, which track how people spend their time in the United States and internationally. It then surveyed 3,995 respondents, whom the company describes as demographically representative, about where they would most value help from a personal assistant. The resulting task areas include caregiving, health, education, household administration, personal finances, shopping and government services.
The examples on the page show how broad that remit is. One asks an assistant to compare kindergartens after a move; another asks it to organize flood-preparation volunteers for four neighbors. Other examples include finding a suitable home in Luzerne County and setting up a way for a grandmother and grandchild to watch a game remotely.
GPT-6 Astra posts several perfect task scores
OpenAI’s results table labels its figures “Best recorded GPT-6 Astra score.” It lists 100% for tasks including booking a car inspection before a road trip, ordering halal groceries within a budget and itemizing the order in a spreadsheet, and unsubscribing from email lists. The same table reports 100% for finding a sensory class that fits a toddler’s needs and family schedule, replacing a lost engagement ring using a photo and diamond report, and finding a pocket-sized device to replace three gadgets.
Other task examples on the page carry lower percentages: 93.1% for shortlisting kindergartens, 98.0% for organizing flood-preparation volunteers, 96.3% for arranging remote game-watching and 90.4% for preparing a meal plan for a parent with a new diabetes diagnosis. OpenAI does not identify the model behind each of those individual examples in the text shown. The figures therefore should not be read as a direct comparison between models or as a single overall LifeVal score.
The public results leave key questions open
The announcement describes the sources used to choose tasks and offers sample outputs, but it does not specify the evaluation’s total task count, how each percentage is calculated or what evidence qualifies as completing a task. It also does not present a baseline or a side-by-side comparison with other models. Those omissions limit what readers can infer from the high scores: a perfect result on a listed workflow does not, by itself, show how consistently a model handles the full range of household and administrative work.
OpenAI says its models have become faster and more capable at browser use, computer use and connectors over the past year, and argues that these abilities extend ChatGPT beyond conversation toward an assistant that can act. LifeVal gives the company a way to describe progress against tasks people say they want help with. For now, its published examples make the ambition concrete, while the measurement details needed to independently assess those percentages remain absent from the announcement.