Everyone has been asking me why Ramp is building a router
We started by helping companies manage spend.
As more of their money goes to buying intelligence, our job is to help 85k customers decide between models, too
Excited about the progress so far, but still a lot of work ahead
We started by helping companies manage spend.
As more of their money goes to buying intelligence, our job is to help 85k customers decide between models, too
Excited about the progress so far, but still a lot of work ahead
How do model routers compare to individual models on real agentic tasks? Introducing the Vals Router Index.
Routers promise to balance quality and cost across a fast-moving model frontier. We put that promise to the test with an independent evaluation of five routers (OpenRouter Router, Jev, Fugu Max) against every single model on the Vals Index, using the same agents, ~1,250 tasks and graders across finance, coding, law and tax.
We found that routing didn’t consistently deliver better value. Every router was beaten on both score and cost by at least one single model.
1
4
66
3,896
Ramp Router is literally free money if you're spending 5.5-8% on top of list prices with OpenRouter
I'll personally come to your offices in NYC or SF to get you onboarded
I'll personally come to your offices in NYC or SF to get you onboarded
Rolling something out to an entire engineering org means dealing with the less fun stuff: budgets, billing, data controls, and more.
We just shipped a bunch to make it easier for Admins to bring teams onboard:
1. Org management: Invite engineers and assign co-admins. Everyone a one-line install for their coding agents, and routing strategies for what they're building.
2. Policies: Configure which models and providers people can use, and set data retention policy at the org and individual levels.
3. Spend controls: Set budget caps for the org, each individual, and each API key. See spending across the org.
4. Key management: See everyone’s key, share provider keys across the org, let personal keys take precedence, and control fallback to Ramp's shared key.
5. Payments: Prepay with a card or pay monthly based on usage.
6. One Ramp account, one bill: Already use Ramp? Sign in with your existing Ramp account and put Router usage on your monthly Ramp statement.
We want every engineer to have the best setup for their work, without making finance and IT nervous about what it's all going to cost.
10
6
167
13,893
Incredible work from @kir_lev @alexosmonov and team to build a model agnostic, cloud native harness that powers all of Ramp’s agents
This is an incredibly high leverage piece of our infrastructure that allows us to move quickly without any vendor lock in
Highly recommend reading the deep dive
This is an incredibly high leverage piece of our infrastructure that allows us to move quickly without any vendor lock in
Highly recommend reading the deep dive
We at Ramp tried a bunch of agent harnesses and ended up building our own. Cloud-native, model-agnostic, tailored to our infrastructure and product.
And we published how we built it: builders.ramp.com/post/building-…
1
1
43
2,909
Eric, Karim, and Zach recruited me to Ramp with a simple pitch to build "Paribus for businesses"
I remember taking the call from my dorm room, wondering if I was ready to leave Stanford for NYC when all my friends were staying in SF
Seven years later, I'm proud of what we've built and even more grateful that the people I've built it with have become family
I remember taking the call from my dorm room, wondering if I was ready to leave Stanford for NYC when all my friends were staying in SF
Seven years later, I'm proud of what we've built and even more grateful that the people I've built it with have become family
Today, Ramp raised $1.85B at a $60B valuation.
Seven years ago, Karim and I wrote our mission on a Post-it: “Save companies time and money.”
Our mission hasn't changed. It’s just gotten bigger.
15
3
296
9,999
Why I use Glass over Codex and Claude Code:
- model agnostic (try the nth new model launch without changing your setup)
- cloud first (vs. frankensteined local vs. cloud distinction)
- built around shared company context (not one person slinging slop grenades back and forth)
And for finance/IT admins:
- govern who can use which models
- see and manage token spend easily
- shared outputs: do the work once, instead of running redundant agents across your org
- model agnostic (try the nth new model launch without changing your setup)
- cloud first (vs. frankensteined local vs. cloud distinction)
- built around shared company context (not one person slinging slop grenades back and forth)
And for finance/IT admins:
- govern who can use which models
- see and manage token spend easily
- shared outputs: do the work once, instead of running redundant agents across your org
17
12
256
26,111
Worth listening
@max11free and @16SamBuck have built an incredible sales team from the earliest days. Feel lucky to learn from them every day.
Talent density deserves just as much attention in sales as it gets in engineering
@max11free and @16SamBuck have built an incredible sales team from the earliest days. Feel lucky to learn from them every day.
Talent density deserves just as much attention in sales as it gets in engineering
@max11free is the definition of a barrel đź’Ą
Built a sales org full of recovering CFOs/controllers, political correspondents, fighter pilots, professional/olympic athletes, technical founders
It’s hard to grasp how much leverage exists in this system unless you’re in it every day
2
2
62
4,016
Every company applying AI into their product should be racing to become model agnostic across the entire stack (router, harness, application)
Don't let them lock you in
Don't let them lock you in
Grok 4.7. 50% off until October 6.
We partnered with @spacexai to make your token spend go further. Get your API key at router.com.
4
1
74
4,847
Veeral Patel retweeted
I asked people: "where should enterprises go for post-training as a service?"
30 votes: Applied Compute (@ypatil125, @rhythmrg, @lindensli)
13 votes: Trajectory (@rronak_, @MichaelElabd, @QuantumArjun)
10 votes: Prime Intellect (@vincentweisser, @johannes_hage)
7 votes: amurati, @johnschulman2)
5 votes: Fireworks (@lqiao, @dzhulgakov, @pawelg, @jamesr66a, Chenyu Zhao, @divchenko)
4 votes: Baseten (@tuhinone, @amiruci, @saltyph, @defpan, @mudithj, @maxkirkby, @oneill_c)
4 votes: Together AI (@vipulved, @ce_zhang, Chris RĂ©, @tri_dao, @percyliang)
2 votes: Arcanum (Rehan Rupawalla, @danielrupawalla, Krish Thawani, @samir_walji)
2 votes: Belvedir (@thezacharyyu)
2 votes: Trainloop (@jackson_stokes, @mlpierce22)
1 vote each:
- AfterQuery (@CarlosGeorgescu, @spencermateega)
- Bespoke Labs (@madiator, @AlexGDimakis)
- Engram (@dan_biderman, @EyubogluSabri, @realJessyLin)
- Mercor (@BrendanFoody, @adarsh_exe, @suryamidha)
- Radixark (@ying11231, @BanghuaZ)
- River AI (@ibab, @dsoboliev, Ievgen Soboliev, Yaroslav Nazarov)
- Unsloth (@danielhanchen, Michael Han)
People couldn't pick a company they work at or founded. Disclosure: I'm a small investor in AfterQuery, Applied Compute, Arcanum and Trajectory. I didn't vote.
30 votes: Applied Compute (@ypatil125, @rhythmrg, @lindensli)
13 votes: Trajectory (@rronak_, @MichaelElabd, @QuantumArjun)
10 votes: Prime Intellect (@vincentweisser, @johannes_hage)
7 votes: amurati, @johnschulman2)
5 votes: Fireworks (@lqiao, @dzhulgakov, @pawelg, @jamesr66a, Chenyu Zhao, @divchenko)
4 votes: Baseten (@tuhinone, @amiruci, @saltyph, @defpan, @mudithj, @maxkirkby, @oneill_c)
4 votes: Together AI (@vipulved, @ce_zhang, Chris RĂ©, @tri_dao, @percyliang)
2 votes: Arcanum (Rehan Rupawalla, @danielrupawalla, Krish Thawani, @samir_walji)
2 votes: Belvedir (@thezacharyyu)
2 votes: Trainloop (@jackson_stokes, @mlpierce22)
1 vote each:
- AfterQuery (@CarlosGeorgescu, @spencermateega)
- Bespoke Labs (@madiator, @AlexGDimakis)
- Engram (@dan_biderman, @EyubogluSabri, @realJessyLin)
- Mercor (@BrendanFoody, @adarsh_exe, @suryamidha)
- Radixark (@ying11231, @BanghuaZ)
- River AI (@ibab, @dsoboliev, Ievgen Soboliev, Yaroslav Nazarov)
- Unsloth (@danielhanchen, Michael Han)
People couldn't pick a company they work at or founded. Disclosure: I'm a small investor in AfterQuery, Applied Compute, Arcanum and Trajectory. I didn't vote.
14
28
425
55,019
I’m hiring FDEs to work on Ramp Router at AI pilled companies
Current/former founders who can build and sell will be unstoppable here
Ideally you are:
- Obsessed with something to an unreasonable degree
- Brimming with crazy ideas
Someone on our growth team put a Ramp ad . Last night we ran a one-night-only Broadway musical about paying bills
- High agency and actually follow through to execution
Email or DM me if interested
Current/former founders who can build and sell will be unstoppable here
Ideally you are:
- Obsessed with something to an unreasonable degree
- Brimming with crazy ideas
Someone on our growth team put a Ramp ad . Last night we ran a one-night-only Broadway musical about paying bills
- High agency and actually follow through to execution
Email or DM me if interested
26
5
290
26,513
Jev in Router.com
Here's a quick look into how we're using @typesafeai's Jev here at Router.com.
We're still innovating with this a ton here at @tryramp , so follow along!
3
0
77
9,528
Ashwin benchmarked Jev as a replacement for LLM-based reranking in Ramp's accounting product, and the results are very promising
Jev matches current accuracy on GPT-5.6 Luna, but decreases tail latency by 10x to 300ms at 3x lower cost
We're excited to get this into production on as we stop running into rate limits :)
Jev matches current accuracy on GPT-5.6 Luna, but decreases tail latency by 10x to 300ms at 3x lower cost
We're excited to get this into production on as we stop running into rate limits :)
27
15
193
18,027
Neel moved Ramp’s structured receipt extraction from Gemini 3.5 Flash-Lite to DeepSeek-V4.1-Flash:
- 70% cheaper (~$40k/month saved)
- about the same accuracy
- about the same latency
Running through router.com
- 70% cheaper (~$40k/month saved)
- about the same accuracy
- about the same latency
Running through router.com
23
9
429
59,006
DM or email me (veeral@router.com) if your team wants help in migrating your own inference over to OS models
1
0
12
3,432
Companies are likely spending millions subsidizing employees’ personal ai use because oai and anthropic can’t figure out how to build a decent account switcher
19
3
306
14,694
This chart is a Rorschach test for how bullish or bearish you are on the AI trade
New from Ramp data: the latest threat to the AI trade. AI companies' revenues are heavily dependent on a small set of customers. 80% of OpenAI and Anthropic's enterprise revenues come from 1% of their customers, and it's not getting better.
This is a level of concentration risk tware category we track. The companies in the top 1% skew heavily toward the tech sector and AI products and services.
What happens in a market correction? All these companies are highly correlated, and an increasing share of our economy is invested in them. Especially as we approach blockbuster IPOs for OpenAI and Anthropic.
7
2
60
8,721