Ramp launches its own AI model router, called Router
TechCrunch Ram Iyer ● Covered by 5 sources
Ramp launched Router, an AI model routing API that swaps between big models. It’s free through 2026, but it also locks Ramp deeper into the AI inference boom.
Based on reporting by TechCrunch, Ram Iyer — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Ramp is trying a new line of business that looks a lot like a toll booth for AI traffic. On Wednesday evening, the expense-management company launched Router, an API service that lets users and companies switch between large language models without rebuilding their setup every time they want a different one.
The pitch is simple enough: Ramp says it has been using its own router internally for the past three years, and now it’s selling that setup to others. The service is only available in the United States for now. It’s free to use for the rest of 2026, though customers still have to pay the actual inference bills from the model providers, and Ramp is dangling a $26 launch credit. Pricing after next year wasn’t disclosed.
Router is meant to work in a similar way to OpenRouter, though Ramp’s current lineup is much smaller. The service includes access to models from OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai. It also gives customers a set of routing strategies, including one that prefers providers’ flex usage tiers and another that can choose models based on up to three user-set benchmarks. Users can send only the hard stuff to pricier models, or test models without hopping between tools.
Ramp is also wrapping the whole thing in the kind of dashboard finance teams tend to love. The company says users can track token spend, cost, latency, fallback attempts, and other details. There’s also an opt-out data retention policy: inputs, outputs, and tool calls are stored for one year by default, and Ramp says it will strip personally identifiable information before using that material to improve the product.
For Ramp, the bet is obvious. Model routing lets it ride the AI inference wave while slotting neatly beside its existing token-usage monitoring and spend-management tools. And if Router becomes a useful place for model testing, as OpenRouter already is, Ramp could end up with deeper ties to AI labs and inference providers, plus another route into selling its core expense software. The company raised $750 million at a $44 billion valuation in June, which means it has plenty of room to chase a second act.
My take — AI-written commentary, not fact-checked reporting
This is the kind of move that makes sense right up until every finance software company decides it is also an AI plumbing company. Ramp already has the billing angle, so Router feels less like a moonshot and more like a neat excuse to sit closer to the money. Clever, sure. Also extremely on-brand for the era of turning every hot market into an invoice.
Read more about this at: TechCrunch
