bolt AI and tech · 2 min
GPT‑6 scored 98% at organizing flood volunteers
flash_on In 20 seconds
GPT‑6 scored 80% on general everyday-task prompts
93.1% kindergarten, 98% volunteers, 90.4% diabetic meal plan
OpenAI built an internal tool to test real‑world errands and chores
Surveyed ~4,000 people; about 100 users created task rubrics
based on a report by CNN, OpenAI unveiled a new internal benchmark and research showing its GPT‑6 Astra model improving on everyday errands and chores.
OpenAI reported GPT‑6 scored 80% on general tasks, and 93.1%, 98%, and 90.4% on specific tests. It surveyed around 4,000 people and had about 100 users build test rubrics.
format_quoteAnd I think that’s like the main use case that we’re really excited about.
cnn.com
help What is not known yet
- When these capabilities will be widely available to all consumers
- Exactly how OpenAI will fix specific safety issues
- How these agent features will be priced for businesses
forum Frequently asked questions
How well did GPT‑6 do on everyday tasks?expand_more
OpenAI reported GPT‑6 Astra scored 80% on general performance for everyday-task prompts.
How much did performance improve over the prior model?expand_more
GPT‑5.2 scored 49.3% on similar prompts in December 2025; GPT‑6 scored 80%.
What kinds of tasks were tested?expand_more
Tasks included shopping, budgeting, finding kindergartens, meal planning, apartment hunting and volunteer coordination.
Did real users help design the tests?expand_more
Yes. OpenAI surveyed about 4,000 people and had roughly 100 users write task rubrics and examples.
newspaper Read next
link Sources (1)
- cnn.com — OpenAI says its AI is getting better at doing things humans do, like shopping and meal planning | CNN Business
How we work: only facts that appear in the cited sources; quotes checked word for word; no opinion and no predictions. Written with the help of artificial intelligence and reviewed by automatic rules.