OpenAI's LifeVal scores GPT-6 Astra 80.6% on everyday tasks

OpenAI scored its newest model, GPT-6 Astra, at 80.6% average reward on a benchmark of everyday chores it built itself and has been using internally for about a year. The company published LifeVal on November 2 alongside a chart that walks the same curve from 49.3% on GPT-5.2 up through GPT-5.4 at 73.0%, GPT-5.5 at 77.4%, GPT-5.6 at 80.0% and finally Astra at 80.6%.

LifeVal is drawn from the US Bureau of Labor Statistics' American Time Use Survey and the Multinational Time Use Study, plus a demographically representative survey of 3,995 respondents and about 100 participants who recorded the specific tasks they wanted an assistant to handle. The categories are caregiving, health, education, household administration, personal finances, shopping and government services. Astra's best recorded runs include meal prep for a parent adjusting to a new diabetes diagnosis (90.4%), enrolling in the right kindergarten after moving (93.1%), watching a game with grandma from far away (96.3%) and managing volunteers ahead of a flood (98.0%).

"To be a truly helpful personal assistant, ChatGPT should be able to take actions that actually get things done," OpenAI writes. The pitch against chat-style evals: a model "can complete the required clicks and still produce an unsuitable result."

The task set itself is not public, there is no leaderboard, and all five scores on the chart are OpenAI's, run by OpenAI, on its own graders. The release lands the same week our personal-agents tracker picked up SJTU and NUS pitching 'Avatar Intelligence' for roughly the same job.