---
title: "LifeVal tests how well AI handles everyday tasks"
url: https://blog.vomini.com/blog/openai-lifeval-everyday-tasks
date: 2026-11-03T02:30:00+00:00
updated: 2026-11-04T16:54:22.523820+00:00
author: "Roberto R."
section: "News"
language: en
site: "Vomini"
license: free to quote with attribution
---

# LifeVal tests how well AI handles everyday tasks

> OpenAI published LifeVal, a test of how well its models handle everyday tasks like finding childcare, budgeting and government paperwork. On OpenAI's numbers, the average score rose from 49.3% for GPT-5.2 to 80.6% for…

OpenAI published LifeVal on 2 November, a test of how well its models help with everyday tasks: finding childcare, using government services, building a budget, preparing for school, changing a recipe, ordering groceries. On OpenAI's own numbers, the average score went from 49.3% for GPT-5.2, released in December 2025, to 80.6% for GPT-6 Astra, released in September.

## How OpenAI picked the tasks

OpenAI started from two large studies of how people spend their time, the American Time Use Survey and the Multinational Time Use Study. It then ran its own survey of 3,995 people to find where they would most want help from a personal assistant. The tasks cover caregiving, health, education, household admin, personal finance, shopping and government services.

OpenAI says it used the test over the past year to track and push its models on these tasks.

## What the scores look like

The score is an average reward across tasks, unweighted. OpenAI lists five models:

| Model | Score |
| --- | --- |
| 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% |

The post also shows single tasks. Comparing kindergartens after a move scored 93.1%. Planning meals for a parent with a new diabetes diagnosis scored 90.4%. Finding a home within a budget and a short commute scored 100%.

OpenAI ties the gains to training on browser use, computer use and connectors, the links that let ChatGPT act inside other apps instead of only answering.

## What is not clear yet

The post does not say whether the tasks or the scoring will be published so others can run them. Every number here comes from OpenAI, about OpenAI models. Earlier this year OpenAI released [MentalHealthBench](/blog/openai-mentalhealthbench-open-benchmark) as an open test; LifeVal, so far, is not described that way.

## What to do this week

If you already use ChatGPT for errands like these, pick one task you do every month and give it the whole job, then check every step it took before you rely on it.

## What we think

Measuring AI on the chores people actually do is more useful than another exam score. The jump between GPT-5.2 and GPT-5.4 is large, and the gains since then are small, which suggests the easy part is done. Independent results will matter more than the chart.

## Sources

- [OpenAI, LifeVal](https://openai.com/index/lifeval/), 2 Nov 2026
- [GPT-6 Astra launch](/blog/gpt-6-astra-launch), our coverage

---

Originally published at https://blog.vomini.com/blog/openai-lifeval-everyday-tasks