Ramp launches Ramp Router to cut LLM costs by 30%
Ramp says the OpenAI-compatible endpoint picks cheaper models that still meet a task's quality bar, and it is opening the tool beyond its own customer base.
Entities: Ramp, Ramp Router, Veeral Patel, Eric Glyman, Karim Atiyeh
Ramp says its new Ramp Router can cut LLM costs by more than 30% by testing models on real work and sending each request to the cheapest model that still clears a quality threshold. In launch posts on X, Victor Riparbelli said the system grew out of an internal router built three years ago and already powers AI products for 70,000 customers, while Rahul Goyal said Ramp is juggling 100-plus use cases as model prices, latency and capabilities keep shifting.
Combined views
235.1K
21 posts, first seen 1d ago
Ramp launches Ramp Router to cut LLM costs by 30%
Ramp says the OpenAI-compatible endpoint picks cheaper models that still meet a task's quality bar, and it is opening the tool beyond its own customer base.
Entities: Ramp, Ramp Router, Veeral Patel, Eric Glyman, Karim Atiyeh
Reactions from ranked influencers
21 posts
you could hire an olympic sprinter to walk your dog. it'd go great. it'd also be insane that's basically how most companies run ai right now. a frontier model on every task, including the ones a model a fraction of the price would nail. we built ramp router to fix this for ourselves. it tests every new model on real work, then routes each request to the cheapest one that clears the bar. routing's just the start, it's making 100+ calls per request on what to cache, when to batch, when to reach for something stronger. output is same or better but already cutting costs by 30%+. the alpha group is run-rating 2.75T tokens/month, at 99.99% uptime. excited to open this up more. lots of you asked about this after last week. we're onboarding companies a few at a time, leave your info on the link below from @rahulgs, @vral and the @RampLabs team and we'll reach out, we'd love to work hard to cut your costs and help you ship the way ramp does
Every week, the price-intelligence-latency frontier shifts, and we expect this trend to continue Across 100+ use cases in our product, keeping all up to date with the right model is a challenge - Either we're losing out on intelligence for the dollars we spend, or we're spending too much money for the intelligence the feature needs Ramp Router lets you benefit immediately without rewriting your application. We cut our LLM costs by 30%, while also making our features smarter and faster. We built it for Ramp. Now we’re opening it up to everyone. get access here: https://ramp.com/router?utm_source=X&utm_campaign=rahul
AI routing is itself an optimization problem. The router has to pick the cheapest model that can still do the job with the quality and latency required. Ramp Router is attacking that increasingly difficult problem: The best model choice from yesterday quickly stops working well. Soon all serious AI products will be running stuff like this behind the scenes. They are claiming to cut AI bill by 30% with a Router that chooses the right model for the right task.
Every week, the price-intelligence-latency frontier shifts, and we expect this trend to continue Across 100+ use cases in our product, keeping all up to date with the right model is a challenge - Either we're losing out on intelligence for the dollars we spend, or we're spending too much money for the intelligence the feature needs Ramp Router lets you benefit immediately without rewriting your application. We cut our LLM costs by 30%, while also making our features smarter and faster. We built it for Ramp. Now we’re opening it up to everyone. get access here: https://ramp.com/router?utm_source=X&utm_campaign=rahul
The year is 2028. Chinese open models have commoditized the frontier labs to dust. Ramp is the most valuable AI company in the world. All is right.
Today we’re launching Ramp Router. 3 years ago, we built an internal LLM router at @tryramp that powers AI products for 70,000 customers. Back then it was mostly about saving money. Now it feels obvious: the best model changes constantly. GPT, Claude, Gemini, Grok, Qwen, DeepSeek, Kimi, GLM - prices and capabilities move every week. So we’re opening up access to everyone. One OpenAI-compatible endpoint. The right model for every request. Lower cost without rewriting your app. Reserve access to use it.
Proud of this one. What started as an internal cost-saving hack now routes AI for 70,000 customers and today it's open to everyone. The best model changes every week. Your code doesn't have to.
Today we’re launching Ramp Router. 3 years ago, we built an internal LLM router at @tryramp that powers AI products for 70,000 customers. Back then it was mostly about saving money. Now it feels obvious: the best model changes constantly. GPT, Claude, Gemini, Grok, Qwen, DeepSeek, Kimi, GLM - prices and capabilities move every week. So we’re opening up access to everyone. One OpenAI-compatible endpoint. The right model for every request. Lower cost without rewriting your app. Reserve access to use it.
We cut our own AI bill by 30% with a Router that chooses the right model for the right task. Now you can too.
Today we’re launching Ramp Router. 3 years ago, we built an internal LLM router at @tryramp that powers AI products for 70,000 customers. Back then it was mostly about saving money. Now it feels obvious: the best model changes constantly. GPT, Claude, Gemini, Grok, Qwen, DeepSeek, Kimi, GLM - prices and capabilities move every week. So we’re opening up access to everyone. One OpenAI-compatible endpoint. The right model for every request. Lower cost without rewriting your app. Reserve access to use it.
ramp web services 🤔 more seriously, you could hire an olympic sprinter to walk your dog. it'd go great. it'd also be overkill. but that's basically how most companies run ai right now. a frontier model on every task, including the ones a model a fraction of the price would nail our obsession is helping companies get more done with fewer dollars and hours, so we built ramp router to fix this for ourselves. it tests every new model on real work, then routes each request to the cheapest one that clears the bar. routing's just the start, it's making 100+ calls per request on what to cache, when to batch, when to reach for something stronger output is same or better but already cutting costs by 30%+. the alpha group is run-rating 2.75T tokens/month, at 99.99% uptime. excited to open this up more. lots of founders asked about this after last week. we're onboarding companies a few at a time, drop your info on the link below from @vral, @rahulgs, and the @RampLabs team and we'll reach out asap, we'd love to work hard to cut your costs and help you ship the way ramp does
Today we’re launching Ramp Router. 3 years ago, we built an internal LLM router at @tryramp that powers AI products for 70,000 customers. Back then it was mostly about saving money. Now it feels obvious: the best model changes constantly. GPT, Claude, Gemini, Grok, Qwen, DeepSeek, Kimi, GLM - prices and capabilities move every week. So we’re opening up access to everyone. One OpenAI-compatible endpoint. The right model for every request. Lower cost without rewriting your app. Reserve access to use it.