Image: OpenAI
OpenAI has introduced LifeVal, a new test that measures how well its AI models handle ordinary chores like finding childcare, building a budget, ordering groceries or dealing with government services. The company published the details on November 2 and announced it on X.
For AGI to benefit all of humanity, it should help people with the everyday tasks that matter most to them. So we built LifeVal, a new benchmark shaped by feedback from thousands of ChatGPT users on where they most want help from a personal assistant—from finding childcare to accessing government services to managing personal finances.
— OpenAI (@OpenAI) November 2, 2026
Over the past year, we’ve used LifeVal to measure and improve how well our models handle these everyday tasks.
openai.com/index/lifeval/
A benchmark is a set of tasks used to score AI models so their progress can be compared over time. Most well known AI benchmarks focus on exams, math or coding. OpenAI says LifeVal is meant to measure something closer to daily life.
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 survey of 3,995 people, which it describes as demographically representative, asking where they would most like help from a personal assistant.
The resulting tasks cover caregiving, health, education, household administration, personal finances, shopping and access to government services. OpenAI says it has used LifeVal over the past year to track and improve its models on these jobs.
The headline result is a steady climb. GPT-5.2, released in December 2025, scored 49.3% on what OpenAI calls average reward. GPT-5.4 reached 73.0%, GPT-5.5 hit 77.4% and GPT-5.6 reached 80.0%. GPT-6 Astra, the newest model on the chart, scored 80.6%.
Image: OpenAI
As RR Tech Pulse points out, most of that gain came early. The jump from GPT-5.2 to GPT-5.4 was 23.7 points, while the last step from GPT-5.6 to GPT-6 Astra added just 0.6 points.
OpenAI also says newer models finish the work faster. A second chart compares scores with how long each attempt takes, and OpenAI says GPT-6 is about 23% faster than GPT-5.6 on these tasks.
Image: OpenAI
The page shows sample tasks with scores for its current models. Enrolling a child in the right kindergarten after a move scored 93.1%, organizing volunteers ahead of a flood scored 98.0% and finding a family home in Luzerne County that fits a budget and a commute scored 100%. Setting up a way to watch a game remotely with a grandparent scored 96.3%, and building a meal plan for a parent newly diagnosed with diabetes scored 90.4%.
OpenAI ties the work to a wider goal. The company says that to be a truly helpful personal assistant, ChatGPT should be able to take actions that get things done, not just give advice. It says its models have become faster and more capable at using a browser, a computer and connected apps, which pushes ChatGPT toward being a true agent.
There are limits to keep in mind. OpenAI built LifeVal, picked the tasks and ran the tests itself. Unite.AI and RR Tech Pulse note that it has not published the task list, the number of tasks or how answers are graded, and no models from other companies appear on the charts.
OpenAI has also not said whether LifeVal will be released for others to run at the time of this writing.
Comments