|
FEATURE RELEASE |
Ramp opens its internal AI model router to the public
The service is built on three years of Ramp's own internal tooling and ships with a $26 launch credit, but pricing after 2026 is unannounced.
Ramp, the corporate expense management platform, launched Router on August 20 - a service that lets companies and developers send AI requests to multiple large language models through a single API endpoint, TechCrunch reports. The pitch is straightforward: one integration point instead of separate connections to every model provider a team wants to try or fall back on.
BY THE NUMBERS
|
Aug 20, 2026 PUBLISHED |
Feature WHAT THIS IS |
|
techcrunch.com SOURCE |
TechCrunch notes that Ramp has been running the router internally for its own AI workloads for three years, which gives the product a longer production history than a typical vendor launch. Router currently reaches models from OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai, per the announcement. The routing strategies are where the practical value sits: one strategy steers traffic toward a provider's flex usage tier when latency allows; another lets Router pick a model automatically based on up to three benchmarks a user specifies; a third routes only harder queries to more expensive models. Users also get a dashboard showing token spend, cost, latency, and fallback attempts, TechCrunch says.
For controllers and accounting managers already watching AI inference costs, that dashboard and the benchmark-driven routing are the features worth testing. The service is US-only for now, is free to use through the end of 2026 (users still pay the underlying inference costs themselves), and includes a $26 credit at launch, according to TechCrunch. What Ramp has not announced is what Router will cost in 2027 - a gap worth tracking before building anything load-bearing on top of it.
Router is comparable in concept to OpenRouter, though TechCrunch notes the latter currently offers a larger model catalog. Finance teams evaluating it should treat it as an inference management layer, not a cost guarantee - and should confirm Ramp's data retention terms before routing anything sensitive through the API.
WHAT TO VERIFY
| Confirm 2027 pricing before building production workflows on Router; Ramp has not announced it. | |
| Check Ramp's data retention policy before routing any sensitive financial data through the API. | |
| US-only at launch - verify availability if your team or vendor is outside the United States. |
SPONSORED
QUESTIONS THIS ANSWERS
Which AI models does Ramp Router support?
Per TechCrunch, Router currently offers access to models from OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai.
How much does Ramp Router cost?
TechCrunch reports the routing service is free through the end of 2026. Users still pay the underlying AI model inference costs. Ramp has not announced pricing for 2027.
Is Ramp Router available outside the United States?
No. TechCrunch states the service is only available in the United States at launch.
SOURCES
Where this comes from
Ten minutes, your requirements, a shortlist.
GO DEEPER
Go deeper
Reported from the linked source pages as published on the dates shown. Audit Friendly has not tested the product unless a review is linked.
The modern accounting tech stack: tools, automations and workflows shaping the next generation of finance operations.
Modern Accounting · Audit Friendly Data · Accounting & Finance Jobs
Audit Friendly · accountingstack.ai