Ramp 推出自有 AI 模型路由服务 Router

企业费用管理平台 Ramp 正紧随 Stripe 建立 AI 推理收费关口的步伐。

Ramp 于周三晚间推出了自己的 AI 模型路由服务 Router。该服务让用户和企业能够通过 API 使用并切换多种大语言模型。该公司表示，在过去三年中，它一直在为自身的 AI 使用需求采用其自建的路由器。

该服务目前仅在美国可用。2026 年剩余时间内可免费使用（但用户仍需支付 AI 模型推理成本），并提供 26 美元的启动额度。该公司尚未说明该服务明年的收费标准。

Router 的功能与 OpenRouter 类似，不过后者提供的 AI 模型选择远多于 Ramp 当前的产品。

Router 提供对 OpenAI、Anthropic、DeepSeek、Moonshot、Minimax、Nvidia、xAI 和 Z.ai 模型的访问。它还提供多种“策略”，帮助客户根据自身偏好将 AI 请求路由至相应模型。例如，其中一种策略允许用户优先选择模型提供商的灵活使用层级；另一种则允许 Router 根据用户指定、最多三项基准测试来决定将查询路由至哪个模型。用户还可以选择仅将困难问题路由至价格更高的模型，或无需切换即可轻松测试模型。

用户还可使用仪表板查看 Token 支出、成本、延迟、回退尝试次数及其他详细信息。

值得注意的是，Router 采用可选择退出的数据保留政策：默认情况下，它会将模型输入、输出和工具调用记录保存一年；不过，该公司表示，在利用这些内容改进产品之前，将删除“个人身份识别信息”。

对于 Ramp 而言，进入模型路由业务带来了双重机会：一方面，它可以进入蓬勃发展的 AI 推理市场；另一方面，它可以向现有客户提供与其既有产品自然契合的模型路由服务。Ramp 现有产品包括 AI Token 使用量监控和 Token 支出管理。

此外，如果 Router 能像 OpenRouter 一样成为具有吸引力的模型测试场所，Ramp 还可能与全球 AI 实验室和推理服务提供商建立长期关系。这将有助于 Ramp 获得新客户，并为销售其费用管理产品开辟新的切入点。Ramp 曾于 6 月以 440 亿美元估值 融资 7.5 亿美元。

Corporate expense management platform Ramp is hot on the heels of Stripe in setting up toll houses for AI inference.

Ramp on Wednesday evening launched its own AI model routing service, dubbed Router, that lets users and companies use and switch between various large language models through an API. The company says it has been using the router it built for its own AI usage needs over the past three years.

The service is only available in the United States for now. It’s free to use for the remainder of 2026 (users will still have to pay for AI model inference costs), and it comes with a $26 credit launch offer. The company did not say how much the service will cost next year.

Router functions similarly to OpenRouter, though the latter offers many more AI model options than Ramp’s current offerings.

Router offers access to models from OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai. It also provides several “strategies” to help its customers route AI requests to models based on their preferences. For example, one lets users set a preference for model providers’ flex usage tiers, while another lets Router choose which model to route queries to based on up to three user-specified benchmarks. Users can also choose to route only difficult problems to expensive models or test models easily without having to switch.

Users get a dashboard, too, that lets them see token spend, cost, latency, fallback attempts, and other details.

Notably, Router has an opt-out data retention policy: It will record model inputs, outputs, and tool calls for one year by default, though the company says it will remove “personally identifiable information before using that content to improve the product.”

For Ramp, entering the model routing business offers a two-pronged opportunity: It gets to tap the booming AI inference market and offer its existing clients a model routing service that fits in neatly with its existing products, which includes AI token usage monitoring and token spend management.

And, if Router proves as attractive of a model testing arena as OpenRouter has, Ramp may also be able to build long-standing relationships with AI labs and inference providers worldwide. That could help Ramp, which raised $750 million at a $44 billion valuation in June, gain new customers and a new point of entry for selling its expense management products.