Ramp launched Router, an AI model routing service that lets developers access and switch between large language models from multiple providers through a single API, the company said.
Router supports models from OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI and Z.ai, and offers routing strategies based on cost, latency benchmarks or usage-tier preferences, according to Ramp. A dashboard shows token spend, cost, latency and fallback attempts across providers.
Ramp said it built and used the router internally for three years before opening it up. The service is free through the end of 2026, plus a $26 launch credit, though customers still pay each model provider's own inference costs; it is currently limited to U.S. users, the company said. Ramp keeps data for one year by default, with personal information removed before any use in product improvement, according to the company.
The launch follows Ramp's $750 million raise in June at a $44 billion valuation, and puts it in direct competition with dedicated routing services like OpenRouter, though Ramp currently supports fewer models, TechCrunch reported.
A payments company shipping model-routing infrastructure is another sign that the aggregation layer between apps and model providers is becoming table stakes rather than a standalone business, and that fintechs already holding a company's spend data have an obvious wedge into it.