OpenAI published LifeVal on November 2, 2026, a new benchmark designed to measure progress of artificial intelligence models on everyday tasks that people actually want help solving.
Unlike traditional academic tests, LifeVal focuses on practical actions: finding childcare, accessing government services, creating budgets, preparing meals, shopping and organizing volunteers, among others. The company states it used the task set over the past year to track and improve the performance of its models.
How LifeVal was built
OpenAI started from two large time-use studies — the American Time Use Survey and the Multinational Time Use Study — and supplemented them with its own demographically representative survey of 3,995 people. The goal was to identify areas where people most wanted help from a personal assistant. The resulting tasks cover caregiving, health, education, household administration, personal finances, shopping and access to public services.
GPT-6 Astra performance
According to OpenAI, GPT-6 Astra (released in September 2026) achieved an average reward of 80.6% on LifeVal, compared to 49.3% for GPT-5.2 from December 2025. In some published examples:
- Enroll in the right kindergarten after moving: 93.1%
- Manage volunteers ahead of a flood: 98.0%
- Find a new home for the family: 100%
- Set up remote game watching with grandma: 96.3%
- Meal prep for a parent with a new diabetes diagnosis: 90.4%
The company highlights that models have become faster, more capable and more aligned on browser use, computer use and connectors, helping ChatGPT move beyond a simple chat tool toward an agentic personal assistant.
Why this matters
For OpenAI, measuring progress on everyday tasks is essential for AGI to broadly benefit humanity. LifeVal represents a step in that direction, although the scores come from the company's own internal evaluation of its models, without the task set and rubrics publicly available for independent reproduction.
Sources
Transparency: This content was created, edited or reviewed with the aid of artificial intelligence. Information was cross-checked with public posts on X and sources available on the internet. Consult the original sources to verify the full context.
By GeekikiBot
