Ramp has launched Router, an AI model routing service that lets users and companies access and switch among large language models through an application programming interface, according to TechCrunch. A Hacker News front-page item listed the headline “Ramp Launches a Model Router” and linked to router.com. TechCrunch reports that Ramp launched the service on Wednesday evening and says the company has used the router internally for its own AI needs for the past three years. The move puts the corporate expense management company into the infrastructure layer around AI inference: not building the models themselves, but helping customers decide which model to call, when, and at what cost. The service is currently available only in the United States, according to TechCrunch. Ramp is making Router free to use for the rest of 2026, though users still pay the underlying AI model inference costs. The launch also includes a $26 credit offer, and Ramp has not said what Router will cost next year. Router currently provides access to models from OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai, TechCrunch reports. The publication describes its function as similar to OpenRouter, while noting that OpenRouter offers many more AI model options than Ramp’s current lineup. The product includes several routing “strategies” that let customers send AI requests based on preferences. TechCrunch cites examples including a preference for model providers’ flex usage tiers, routing decisions based on up to three user-specified benchmarks, sending only difficult problems to more expensive models, and testing models without changing integrations. Ramp is also offering a dashboard for operational visibility. According to TechCrunch, users can track token spend, cost, latency, fallback attempts, and other details. That positioning fits with Ramp’s existing products around AI token usage monitoring and token spend management, as described in the report. The data policy is a notable implementation detail. TechCrunch reports that Router has opt-out data retention: by default, it records model inputs, outputs, and tool calls for one year. Ramp says it removes personally identifiable information before using that content to improve the product, according to the report. For Ramp, TechCrunch frames the launch as a way to participate in the AI inference market while giving its existing clients another service adjacent to expense management. The report also notes that Ramp raised $750 million at a $44 billion valuation in June, giving the company room to extend into products that monitor, route, and manage AI spend. Who benefits: Teams experimenting across multiple large language model providers could benefit from a single API, routing rules, and a dashboard for spend and latency. Ramp could also benefit if Router becomes an entry point for customers that later adopt its broader expense-management products. Who's exposed: Customers with strict data-handling requirements will need to examine Router’s default one-year retention policy and opt-out mechanics before adopting it.