这个模型并不像其他 AI 巨头那样「刷分」，而是朝着小型化、端侧化、低延迟的方向做了极致优化，将视觉处理所需的 Token 降到传统 ViT 的 1/16，极大降低延迟，可以根据摄像头捕捉到的内容实时给出判断，反应速度非常快。
公开资料显示，全球已有超3000万个智能体在为用户提供服务，助力全球日均Token消耗量过去两年增长近300倍。
，这一点在新收录的资料中也有详细论述
The legal argument was not proven out. Despite our legal argument about copyright protection, the ongoing litigation kept the question open. Organizations like Forbes and the SF Standard who had shown interest in our model couldn’t afford to bet on an unresolved legal argument. Without regulation clarity, Tess couldn’t compete with models like NanoBanana, DALLE, and Midjourney.，这一点在新收录的资料中也有详细论述
Андрей Шеньшаков