人工智慧产业传统上将模型分类为开放（允许免费修改与重新发布）或封闭（具有严格的存取控制，如 OpenAI 与 Anthropic 的模型）。最近，中国人工智慧实验室 Alibaba 与 Moonshot 引入了基于收益分润模式的具战略性的第三种类别。此方法特别针对主要云端服务供应商，包含 Amazon、Microsoft 与 Google，要求他们为尖端模型（如 Kimi K3）支付存取费用。

虽然这些授权允许新创公司与个人开发者以几乎零成本下载并自订模型，但显著的财务门槛会触发付款义务。具体而言，如果模型即服务企业在连续 12 个月内使用 Alibaba 的 Qwen3.8-Max 创造超过 5000 万美元的收益，或使用 Moonshot 的 Kimi K3 创造超过 2000 万美元的收益，该公司便需要支付费用。这种复杂的商业策略利用了他们最初以极低成本广泛提供模型的策略。

这种收益分润方法面临潜在风险，例如来自华盛顿的监管打击，或者云端公司可能干脆拒绝支付并完全放弃这些模型。尽管如此，像 Qwen 与 Kimi 这类中国人工智慧模型的高能力与显著较低的成本展现了实质的市场竞争力。这股趋势对 OpenAI 与 Anthropic 等老牌美国企业构成了相当大的威胁，因为他们正试图在预期的公开上市前巩固其企业估值。

The AI industry traditionally classified models as either open, allowing free modification and redistribution, or closed, featuring strict access controls like those from OpenAI and Anthropic. Recently, Chinese AI laboratories Alibaba and Moonshot introduced a strategic third category based on a revenue-share model. This approach specifically targets major cloud service providers, including Amazon, Microsoft, and Google, requiring them to pay access fees for cutting-edge models such as Kimi K3.

While these licenses permit startups and individual developers to download and customize the models at virtually no cost, significant financial thresholds trigger payment obligations. Specifically, if a model-as-a-service business generates over $50 million in revenue using Alibaba’s Qwen3.8-Max, or exceeds $20 million in revenue utilizing Moonshot’s Kimi K3 within a span of 12 consecutive months, the company is required to pay fees. This sophisticated tactic capitalizes on their initial strategy of offering models widely for minimal costs.

This revenue-share approach faces potential risks, such as regulatory crackdowns from Washington or the possibility that cloud companies might simply refuse to pay and drop the models entirely. Nevertheless, the high capability and significantly lower costs of Chinese AI models like Qwen and Kimi present substantial market competition. This trend poses a considerable threat to established US firms like OpenAI and Anthropic as they attempt to solidify their enterprise valuation ahead of anticipated public listings.