OpenAI releases LifeVal benchmark for everyday personal assistant tasks

OpenAI has released LifeVal, a benchmark for everyday personal assistant tasks such as finding childcare, accessing government services and managing personal finances.

The benchmark was built from feedback from thousands of ChatGPT users, and OpenAI says it has used LifeVal over the past year to measure and improve how its models handle these tasks. OpenAI's own posts give no scores or models tested.

Rohan Paul, on X, describes the benchmark as covering caregiving, health, education, finance, shopping and household admin tasks, such as booking a car inspection or ordering halal groceries within budget. Paul says a chart shows GPT-6 Astra scoring 80.6%, up from 49.3% for GPT-5.2 about nine months earlier. These scores come from Paul's description of the chart, not from an OpenAI statement.