OpenAI has published LifeVal, an evaluation it says measures “AI progress on the tasks that matter in everyday life”, and its chart puts its newest model, GPT-6 Astra, at an average reward of 80.6%. The post is dated 2 November.
On OpenAI’s chart, GPT-5.2 scored 49.3% and GPT-5.4 scored 73.0%. By our arithmetic, that one release added 23.7 points, while the three releases since added 7.6 points combined, and GPT-6 Astra sits 0.6 points above GPT-5.6.
Animated chart. Chart: FourWeekMBA.
The key insight:OpenAI says it built LifeVal from where people want help in daily life, and its chart shows most of the measured gain arriving in one release, with the three releases since adding 7.6 points combined, by our arithmetic.
OpenAI says it “built an evaluation of tasks covering areas where users wanted the most help from AI and personal agents” and used it “to track and drive improvement in modelperformance on everyday tasks over the past year”.
The tasks start from two time-use studies, the American Time Use Survey and the Multinational Time Use Study, OpenAI says. It added “a new demographically representative survey of 3,995 respondents” to find where people most wanted help from a personal assistant.
The resulting tasks, according to OpenAI, cover caregiving, health, education, household administration, personal finances, shopping and accessing government services. The post says ChatGPT has “more than one billion people” using it every week.
Average reward (unweighted) on OpenAI’s LifeVal everyday-task evaluation by model release, as charted by OpenAI on 2 November 2026: GPT-5.2 49.3%, GPT-5.4 73.0%, GPT-5.5 77.4%, GPT-5.6 80.0%, GPT-6 Astra 80.6%; point gains are our arithmetic.
The Scores by Model
OpenAI’s chart plots “Average reward (unweighted)” against model release date. It shows GPT-5.2 at 49.3%, GPT-5.4 at 73.0%, GPT-5.5 at 77.4%, GPT-5.6 at 80.0% and GPT-6 Astra at 80.6%, across release dates from 11 December 2025 to 3 September 2026.
By our arithmetic, the total gain from GPT-5.2 to GPT-6 Astra is 31.3 points. About three quarters of it, 23.7 points, came between GPT-5.2 and GPT-5.4.
Example Tasks and Scores
OpenAI lists example workflows with scores for its current models. Enrolling in the right kindergarten after moving scored 93.1%, managing volunteers ahead of a flood 98.0% and finding a new home for a family 100%, the post shows.
Watching a game with grandma from far away scored 96.3%, and meal prep for a parent adjusting to a new diabetes diagnosis 90.4%. A table of “Best recorded” GPT-6 Astra scores lists 100% on six tasks, including ordering halal groceries within a budget and unsubscribing from email lists.
OpenAI says it has trained its models to become “much faster, more capable, and more aligned on browser use, computer use, and connectors tasks”, moving ChatGPT “towards a truly agentic personal assistant”.
Business Pill · The Model and the Harness
A Business Pill on how the model and the harness around it split the work of an AI agent.
As we read it, LifeVal changes what OpenAI reports: the post says it goes beyond academic assessments of model knowledge to measure how well models get everyday tasks done.
As we read the chart, the slower gains since GPT-5.4 can reflect an evaluation nearing its ceiling as much as slower model progress; the page does not say which, and it does not mention any runs by outside groups.
OpenAI, LifeVal, 2 Nov 2026
“To be a truly helpful personal assistant, ChatGPT should be able to take actions that actually get things done”
Three Implications
FOR CHATGPT USERSOpenAI says the tasks cover caregiving, health, education, household administration, personal finances, shopping and accessing government services.
FOR AGENT BUILDERSOpenAI says it trained its models on browser use, computer use and connectors tasks, and lists 100% scores on six tasks such as booking a car inspection.
FOR BENCHMARK READERS The page gives an unweighted average reward but not the task count or grading method, so the 80.6% cannot yet be compared with outside tests.
The framework reads: “The first is what your underlying technology can do. The second is what your users can actually do with it.”
As we read it, LifeVal is OpenAI trying to put a number on the second curve, scoring models on tasks people say they want done rather than on knowledge tests. The framework was published in June 2026 and does not discuss LifeVal.
LifeVal is OpenAI’s own evaluation of its own models, and its scores have not been independently verified. The post does not give the number of tasks, explain how answers are graded beyond the average-reward label, compare models from other companies or link to the task set. The example scores are for selected tasks, not the full evaluation. We did not contact OpenAI.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
OpenAI says its newest model, GPT-6 Astra, averages 80.6% on LifeVal, its new test of everyday-life tasks, up from 49.3% for GPT-5.2; by our arithmetic, 23.7 of those 31.3 points came in a single release.
A note on sourcing. Scores, task examples and survey details are from OpenAI’s LifeVal post of 2 November 2026; model scores are read from its chart. LifeVal is OpenAI’s evaluation of its own models. Point gains are our arithmetic, and we did not contact OpenAI.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
