Technology and Science20 August, 20262 min read
Ramp Launches Router.com AI Model Router in the United States
This article is an automated aggregation of coverage from news outlets. Our editorials offer in-depth analysis curated by the NewsCord team.
Router.com provides a single API to access and route among OpenAI, Anthropic, DeepSeek, Moonshot models Routes each request to the lowest-cost model meeting required performance, cutting bills
Ramp’s Router goes live
Ramp launched Router, an AI model routing service that lets users and companies access and switch between various large language models through a single API, positioning the corporate spend management platform in the AI inference market. The service is only available in the United States and is free for the remainder of 2026, though users still pay the inference costs charged by the models they use. Ramp says it developed and used the router internally for its own AI needs over the past three years, and TechCrunch reported the product was presented the night of Wednesday.
Router connects models including OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai, and it offers a dashboard to show token spend, cost, latency, and fallback attempts. In its launch framing, Ramp described Router as a layer that routes each request to the lowest-cost model that meets the required performance level, while also providing automatic fallback when a provider fails.
Strategies, benchmarks, and retention
Router includes multiple routing strategies, including one that prioritizes providers’ flex usage tiers and another that lets Router choose destinations based on up to three user-specified benchmarks. Ramp also offers an option to route only difficult problems to expensive models, while another approach is described as enabling teams to test models without switching their production setup. The service records model inputs, outputs, and tool calls for one year by default under an opt-out data retention policy, with Ramp saying it will remove “personally identifiable information before using that content to improve the product.”
TechCrunch said Router functions similarly to how OpenRouter operates, but noted that OpenRouter offers many more AI model options than Ramp’s current offerings. In the launch materials, Ramp’s chief technology officer Rahul Sengottuvelu said, "Router puts every token in one place and sends each request to the model that delivers the right performance at the right cost."
Cost claims and what’s next
Ramp’s launch claims that customers already using Router have reduced their inference costs by 40% on average, and it also says it cut its own inference costs by approximately 30% for the same output while delivering 99.9%+ reliability across production traffic. The company ties Router’s routing decisions to Ramp SWE-Bench, described as a benchmark built from production engineering tasks, and PR Newswire says Router continuously tests new models against real Ramp engineering work. In addition to the cost focus, Router is presented as connecting model selection decisions to Ramp’s broader AI spend visibility and controls, giving companies a clearer view of what AI work costs, who owns it, and where to optimize.
“Customers already using Router have reduced their inference costs by 40% on average.”
PR Newswire said AI spend has grown 20.7x since June 2025, and it also stated Router is live now at router.com with free routing through 2026 and $26 in free credits for new users. TechCrunch reported that Ramp did not say how much the service will cost next year, leaving the post-2026 pricing structure an open question for teams evaluating the switching cost.