Editorial illustration for Ramp launches AI model router with one-year data retention by default
Ramp Launches AI Model Router with Data Retention
Ramp built its own AI routing tool three years ago just to manage the company's internal model usage. Now it's selling that tool to everyone else. The corporate expense management company launched Router on Wednesday evening, an API service that lets businesses switch between large language models from OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai without rebuilding their integrations each time.
The move puts Ramp in the same lane as Stripe, another payments company that's found a way to tax the AI inference boom rather than just process card swipes. Router is free through the end of 2026, though customers still pay for the underlying model costs, and new users get a $26 credit to start. Ramp hasn't said what it will charge once the free period ends.
The service works a lot like OpenRouter, minus the sheer number of model options, but adds features aimed at cost control: routing strategies based on provider flex tiers, custom benchmarks, or problem difficulty, plus a dashboard tracking token spend, latency, and fallback attempts. It's U.S.-only for now. One detail buried in the launch drew immediate attention.
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.
Why this matters
Ramp built Router for its own workloads over three years, which gives it a credibility edge Stripe's newer effort doesn't have yet. But the default one-year retention window deserves scrutiny from anyone routing production traffic through it. "Opt-out" means the data collection is the default state, not something teams have to actively enable, and Ramp's promise to strip PII "before using that content to improve the product" leaves open exactly what counts as PII and what other uses that content might serve in the meantime.
For developers weighing model routers, the free pricing through the end of the year is a real incentive, but free-through-2025 also signals this is a land-grab phase where the business model isn't settled. Founders building on top of Router should read the retention terms line by line rather than assume defaults will match their compliance needs. The bigger pattern worth watching: expense software and payments companies are now positioning themselves as AI infrastructure gatekeepers, and the data policies they set here will shape norms across the industry.
Related Reading
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
From Archives Meta AI’s New Mac App Can See Your Screen and Answer Questions
From Archives AI Pioneer Warns Synthetic Data Can't Capture Human Minds
Common Questions Answered
What AI models can be routed through Ramp's Router service?
Ramp's Router API allows businesses to switch between large language models from OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai without requiring teams to rebuild their integrations each time. This multi-model support enables companies to optimize their AI infrastructure by easily switching between providers based on their needs.
How long does Ramp retain data by default in Router?
Ramp's Router service has a one-year data retention window set as the default for all users. This opt-out data collection approach means teams must actively disable retention rather than enable it, which has raised concerns about privacy practices and what qualifies as personally identifiable information.
What complementary products does Ramp offer alongside Router?
In addition to Router, Ramp provides AI token usage monitoring and token spend management tools that integrate with its model routing service. These products work together to help corporate expense management clients track and optimize their AI infrastructure costs across multiple language models.
Why did Ramp develop Router and what competitive advantage does this give them?
Ramp built Router internally three years ago to manage its own model usage before deciding to commercialize it. This three-year track record of real-world deployment gives Ramp a credibility edge over competitors like Stripe, whose model routing efforts are newer and lack this proven production experience.
What market opportunity does Router represent for Ramp?
Entering the model routing business offers Ramp a two-pronged opportunity to tap into the booming AI inference market while offering its existing corporate clients a model routing service that integrates seamlessly with its current expense management and AI monitoring products. This positions Ramp to capture additional revenue from businesses increasingly relying on multiple large language models.
Further Reading
- Ramp launches its own AI model router, called Router - Yahoo Tech
- Ramp launches Ramp Router, an AI model routing product built on three years of internal use - TBPN Digest
- Ramp launches public LLM router to cut infrastructure costs - Digg
- /C O R R E C T I O N -- Ramp/ - Yahoo Finance
- Ramp opens Router.com Model Gateway With Free Routing Through 2026 - Unite.AI
