← All benchmarks

LifeVal

Agents · 2026-11-02

OpenAI's LifeVal: 169 everyday-life tasks chosen from the American Time Use Survey, the Multinational Time Use Study and a 3,995-person survey of where people want help, spanning caregiving, health, education, household administration, personal finances, shopping and government services. Tasks ask the model to get things done rather than advise, such as shortlisting kindergartens or adding a deed to a county records cart, and OpenAI ties its progress to browser-use, computer-use and connector skills. A task's attempts are averaged into a task mean and the score is the unweighted mean of the 169 task means (reward × 100), without willingness-to-pay weighting.

Benchmark source: OpenAI

Top models (higher is better)