OpenAI · 2026-11-02 · notable
LifeVal — OpenAI's test of AI help with everyday life tasks
LifeVal is OpenAI's new internal test of everyday tasks such as finding childcare, budgeting and government paperwork. OpenAI says GPT-6 Astra scores 80% overall, up from 49.3% for GPT-5.2.
OpenAI now measures how well its models get real errands done, not just how much they know.
Key specs
| Gpt 6 astra | 80% |
|---|---|
| Gpt 5.2 | 49.3% |
What is it?
LifeVal is an OpenAI evaluation of practical everyday tasks where people most want help from AI, announced on 2 November 2026. Its tasks cover caregiving, health, education, household administration, personal finances, shopping and access to government services. OpenAI says it used LifeVal over the past year to track and drive improvement on these tasks.
How does it work?
OpenAI started from two large studies of how people spend their time, the American Time Use Survey and the Multinational Time Use Study, then ran its own representative survey of 3,995 people to find where they most wanted an assistant. About 100 users with proven expertise wrote tasks, the personal context they would share, and a rubric for judging success. Example scores for current models include 93.1% on picking a kindergarten after a move and 90.4% on a meal plan for a parent with new diabetes.
Why does it matter?
Most well-known benchmarks test knowledge or coding, while LifeVal scores the website navigation, use of chat history and judgment a personal agent needs to finish errands. It complements GDPval, OpenAI's work-task evaluation. LifeVal is internal: the tasks and data are not public, so outside groups cannot reproduce the scores.
Who is it for?
people tracking AI agents and how labs measure them
Try it
Read OpenAI's post: openai.com/index/lifeval