AI 导读
企业正撞上 token 成本天花板： AT&T 通过将任务路由到更便宜的模型，将 AI 任务成本削减了 56%，质量仅损失 2%；而 Uber 在 4 个月内烧光了全年 AI 预算。 这推动企业 AI 转向路由、固定费率访问和自有/开源模型，因为在规模化场景下，每个请求都用最强的专有模型在经济上并不合理。
正文
Enterprises are hitting a token-cost ceiling:
AT&T cut AI task costs 56% with only a 2% quality loss by routing work to cheaper models, while Uber burned its entire annual AI budget in 4 months.
That pushes enterprise AI toward routing, flat-rate access, and owned/open models, because using the strongest proprietary model for every request does not make economic sense at scale.
来源：@rohanpaul_ai · x.com