LifeVal | OpenAI
November 2, 2026
Measuring AI progress on tasks that matter in everyday life.
The LifeVal video showcases everyday tasks, including shopping, caregiving, moving, and reading with children.
OpenAI’s mission is to ensure that artificial general intelligence (AGI) benefits all of humanity. For the more than one billion people who now use ChatGPT each week, we want to build AI that helps navigate real everyday challenges—such as finding childcare, accessing government services, creating budgets, prepping for school, adapting recipes, ordering groceries, and designing workout routines.
When people turn to AI to make their lives easier, they need more than just advice or chat responses. To be a truly helpful personal assistant, ChatGPT needs to take real action to get things done.
To measure model performance on everyday tasks over time, we built a benchmark evaluation covering the areas where users most want help from AI and personal agents. Over the past year, we have used this evaluation to track and drive improvements in our models’ day-to-day performance.
Model performance on everyday tasks over time
Background
To identify the tasks that matter most to people, we looked to two large-scale studies tracking time use across US and international populations: the American Time Use Survey and the Multinational Time Use Study. Together, these datasets provide a comprehensive inventory of everyday tasks across demographics, times of day, and seasons. To better understand where people value AI assistance the most, we built on these existing surveys with a new study of 3,995 demographically representative respondents. For each task, we asked questions designed to surface where everyday people most want support from a personal assistant. The resulting tasks span key areas where people want help the most, including caregiving, health, education, household management, personal finance, shopping, and navigating government services.
Individual Task Examples
Here are several examples of how our current models perform across real everyday workflows.
Enrolling in the Right Kindergarten After Moving
93.1%

Compare kindergartens near our new home and put together a shortlist of the best options.
I’ve narrowed it down to three kindergartens and compiled the enrollment details. There are two preferences left for you to decide on. Would you like to review the shortlist?
Organizing Volunteers Ahead of Flooding
98.0%

Help me organize a community flood-response volunteer team to assist four neighbors who need flood-proofing support.
I’ve prepared a brief action plan to help volunteers assess each location and get ready for the incoming severe weather.
Finding a New Home for My Family
100%

Find a home in Luzerne County within my budget that has an accessory dwelling unit and a short commute after school drop-off. Add the property deed records to my cart so I can review them.
I found a home that meets your criteria. I’ve calculated the realty transfer tax and added the deed to your county records cart for your review.
Watching the Game with Grandma from Far Away
96.3%

Set up a simple way for my grandma and me to watch the game together remotely.
The easiest watch-together setup is ready, complete with a backup plan if the stream cuts out. Would you like me to walk you and Grandma through it?
Meal Prep for Parents Newly Diagnosed with Diabetes
90.4%

Help my parents adjust to a recent diabetes diagnosis by putting together a practical care and meal plan.
The first week’s plan is ready, including meal plans, reminders, and a list of questions to ask their healthcare team. Want to walk through it together before sharing it with your parents?
Our latest models achieve strong scores across many LifeVal tasks.
| Task | Best Recorded GPT‑6 Astra Score |
|---|---|
| Find and book a sensory class that fits my toddler’s needs and our family schedule | 100% |
| Look up when the car is due for service and schedule an inspection ahead of a road trip | 100% |
| Order halal groceries for my three favorite meals, staying under budget, and itemize the order in a spreadsheet | 100% |
| Find a replacement for my wife’s lost engagement ring using photos and the diamond certification | 100% |
| Find and order a pocket-sized device to replace my phone, e-reader, and handheld console | 100% |
| Unsubscribe from an email list | 100% |
| View more + |
More Task Results
Over the past year, we have trained models to become faster, more capable, and better aligned across browser use, computer use, and connectors tasks. Enabling these capabilities in ChatGPT also expands the types of tasks the model can assist with, moving ChatGPT beyond a simple conversational tool toward a true agentic personal assistant.
Why This Matters
To ensure AGI benefits all of humanity, we need to measure and improve how models perform on the everyday tasks that matter most to people. This means moving beyond academic benchmarks that test a model’s knowledge, and instead measuring how well AI models can assist people with making decisions, managing day-to-day responsibilities, and accomplishing meaningful real-world tasks. We hope LifeVal marks a meaningful step toward that goal.






