OpenAI publishes LifeVal, a benchmark for everyday agent tasks, with GPT-6 Astra at 80.6%OpenAI 发布日常生活智能体基准 LifeVal，GPT-6 Astra 得 80.6%
LifeVal grades agents on caregiving, health, education, household administration, finance, shopping and government services, drawing on time-use studies and a 3,995-person survey.LifeVal 覆盖照护、健康、教育、家务行政、金融、购物与政务服务，任务来源包括时间使用研究与 3995 人调查。
GPT-6 Astra averaged 80.6% reward in OpenAI's own chart, up from 49.3% for GPT-5.2 and 80.0% for GPT-5.6. Cited examples include 93.1% for comparing kindergartens after a move, 98.0% for organising flood volunteers and 90.4% for a first-week diabetes care and meal plan. The task set is not public and only OpenAI models are scored, so these are self-graded figures, not independently reproduced results.
GPT-6 Astra 在 OpenAI 自家图表中平均得 80.6%，高于 GPT-5.2 的 49.3% 与 GPT-5.6 的 80.0%。引用示例包括搬家后比较幼儿园 93.1%、组织防汛志愿者 98.0%、糖尿病首周照护与膳食计划 90.4%。任务集未公开、只给 OpenAI 自家模型打分，因此这些是自评数字，并非独立复现的结果。