Rohan Paul · @rohanpaul_ai · X·2026-09-19 03:08·3小时前
AI 导读

Chamath 撰文指出,开源权重模型与前沿闭源模型的能力差距仅约 4 个月,因此不必赢下所有 benchmark,只需让闭源模型的溢价变得可谈判。对许多日常企业负载而言,更低的成本、更深的定制、基础设施控制权以及把专有数据和工作流留在公司内部,可能比略微落后于前沿更重要。市场将走向混合格局,而非赢家通吃。

Rohan Paul@rohanpaul_ai
38AI 编辑部评分,满分 100
2026-09-19 03:08· 3小时前
AI 导读

Chamath 撰文指出,开源权重模型与前沿闭源模型的能力差距仅约 4 个月,因此不必赢下所有 benchmark,只需让闭源模型的溢价变得可谈判。对许多日常企业负载而言,更低的成本、更深的定制、基础设施控制权以及把专有数据和工作流留在公司内部,可能比略微落后于前沿更重要。市场将走向混合格局,而非赢家通吃。

Open weights do not need to win every benchmark. They just need to make the closed-model premium negotiable.

a very nice article by Chamath.

There's just 4-month capability lag of open-weight models. For many everyday enterprise workloads, being slightly behind the frontier may matter far less than lower costs, deeper customization, control over infrastructure, and keeping proprietary data and workflows inside the company.

we are toward a hybrid market rather than a winner-takes-all outcome.

Chamath Palihapitiyahttps://x.com/i/article/2101002509875892224

来源:Rohan Paul· x.com