Breaking

OpenAI Unveils LifeVal Benchmark, Where GPT-6 Astra Scores 80.6% on Everyday Tasks

November 2, 2026 at 15:02 EST

Foundation Models
Research & Papers

OpenAI has unveiled LifeVal, a new benchmark that tests how well AI systems handle everyday tasks such as childcare, budgeting and interacting with government services, with its GPT-6 Astra model scoring 80.6%.

OpenAI · LifeVal benchmark

OpenAI's GPT-6 Astra scores 80.6% on a benchmark of everyday tasks

LifeVal tests AI on childcare, household budgeting and government services. The score is up from 49.3% for GPT-5.2, OpenAI's model from a year earlier.

80.6%

GPT-6 Astra score on LifeVal

49.3%

GPT-5.2 score, a year earlier

+31.3 pts

Gain in one year (about 1.6×)

LifeVal score, drawn to scale

Column height is proportional to each score (100% = full height).

49.3%

GPT-5.2
(a year earlier)

80.6%

GPT-6 Astra
(LifeVal)

The newer model's column is about 1.6 times the height of the earlier one, a gain of 31.3 percentage points.

What the score does not yet show

One headline number, without the details needed to interpret it

Construction

Number of tasks and how they were selected

Grading

How answers are scored and the rubric used

Verification

No independent evaluation reported yet

Competitors

No other developers' scores on the same test

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.