OpenAI Tests GPT-6 Astra on Shopping and Meal Planning

OpenAI introduced an internal benchmark that measures how AI models handle shopping, budgeting, meal planning and other everyday tasks, according to CNN. The company said GPT-6 Astra scored 80% overall, compared with 49.3% for GPT-5.2.

GPT-6 Astra scored 93.1% on finding a kindergarten after a family moved, 98% on managing community volunteers preparing for a flood and 90.4% on creating a meal plan for a parent recently diagnosed with diabetes.

OpenAI developed the evaluation using surveys of about 4,000 people. A smaller group of about 100 users proposed tasks, supplied relevant personal context and created criteria for judging the results.

The benchmark tests skills including website navigation, reviewing information from chat histories and producing conclusions that match a user's requirements. OpenAI previously released GDPVal to assess model performance on workplace tasks across industries.