OpenAI Launched LifeVal on November 2. The 80.6% Astra Score Going Around Is a Relay; the Page Prints Only Example Cards.
OpenAI published LifeVal on November 2, an evaluation of everyday tasks across caregiving, health, education, household administration, personal finance, shopping, and government services, built from the American Time Use Survey and the Multinational Time Use Study plus a survey of 3,995 respondents. The page shows example cards scored at 93.1%, 98.0%, 100%, 96.3%, and 90.4%. The model-average chart circulating in press, with GPT-6 Astra at 80.6% and a step up from GPT-5.2 at 49.3%, is not printed on OpenAI's page. The eval is self-built, self-graded, and scores only OpenAI models. We read the LifeVal page.
By T.W. Ghost
OpenAI published LifeVal today, November 2, 2026: an evaluation of how well its models handle everyday tasks, the ones time-use surveys say people actually spend their days on. The domains: caregiving, health, education, household administration, personal finances, shopping, and government services. The construction: task categories drawn from the American Time Use Survey and the Multinational Time Use Study, plus a survey of 3,995 respondents about what they do and what they would want help with.
The number traveling with the launch is 80.6%, GPT-6 Astra's average at the highest reasoning effort, at the top of a step chart that starts with GPT-5.2 at 49.3%. That chart is not on OpenAI's page. What the page prints is a set of example cards with individual scores: 93.1%, 98.0%, 100%, 96.3%, and 90.4%. Those are examples, selected by OpenAI, and they read like best cases, not an average. The 80.6% and the rest of the step chart come from secondary write-ups; until OpenAI puts them in a table, they are press. We read the LifeVal page. Here is what held up.
What Is on the Page, and What Is Not
| LifeVal (OpenAI, November 2, 2026) | |
|---|---|
| Domains | Caregiving, health, education, household admin, personal finances, shopping, government services |
| Built from | American Time Use Survey; Multinational Time Use Study; survey of 3,995 respondents |
| Example card scores on the page | 93.1%, 98.0%, 100%, 96.3%, 90.4% |
| Model-average chart | Not on the page; circulating figures (Astra 80.6%, GPT-5.6 80.0%, GPT-5.2 49.3%) are relays |
| Public rubric | Not published |
| Non-OpenAI models | None scored |
| Who grades | OpenAI |
Examples are not an average. Five cards in the ninety-to-hundred range tell you what a good run looks like. They do not tell you what the typical run looks like, and the number that would, the 80.6%, is the one OpenAI did not print.
Self-built, self-graded, single-vendor. LifeVal measures OpenAI models on tasks OpenAI chose with a rubric OpenAI has not released. It may be a good eval. It is not a board.
Keep it apart from Rosalind. OpenAI's other life-adjacent release this week, GPT-Rosalind-6, is a biology access tier priced at a premium. LifeVal is a marketing eval about household tasks. The two share a week and nothing else.
Verified vs Unconfirmed: The Scorecard
| Claim | Verdict |
|---|---|
| LifeVal published Nov 2, 2026; domains as listed | Verified (OpenAI page) |
| Built from ATUS, MTUS, and a 3,995-respondent survey | Verified (OpenAI page) |
| Example cards 93.1%, 98.0%, 100%, 96.3%, 90.4% | Verified as examples (OpenAI page) |
| GPT-6 Astra 80.6%; GPT-5.6 80.0%; GPT-5.2 49.3% | Press (secondary relays; not on the page) |
| Public rubric; third-party models on the board | Unverified (absent) |
What To Do
Cite the page for what LifeVal is and footnote any percentage that is not on it as press.
Do not route on it. An eval with no public rubric and one vendor's models cannot tell you which lab's model does your household admin better.
The frontier is elsewhere: the boards that score more than one vendor were rebuilt last week, and even those carry a private 15%. Our model comparison has the current standings side by side.
Sources
*Not sure whether a vendor's everyday-tasks eval says anything about your tasks? Take the free 2-minute quiz and get matched. Then read about the model at the top of that unpublished chart: GPT-6 Astra.*