智东西4月21日报道，昨夜，Kimi正式发布并开源旗舰模型K2.6，带来其迄今最强的代码能力、长程任务执行和Agent集群能力。 根据官方公布的多项基准测试，Kimi K2.6在博士级难度的“人类最后的考试”（Humanity’s Last Exam）中，K2.6以54.0%的得分位居第一；在评估Agent深度检索能力的DeepSearchQA中，更是以92.5%的高分大幅领先GPT-5.4和Ge ...

AI has shifted from a supportive technology into the core engine of digital product development. In my project experience ...

The Movie Studio level has plenty of great ramps and areas for sweet tricks, but they’re all fairly compartmentalized into different sections of the studio. It can be difficult to chain them all ...

A practical guide to Perplexity Computer: multi-model orchestration, setup and credits, prompting for outcomes, workflows, ...