Ricky | clipur.com
1.8K posts
Ricky | clipur.com
@eric_jacks75737
Digital creator/clipper 🖇️ https://t.co/1lw1Fx5SID
Katılım Kasım 2025
1.8K Takip Edilen1K Takipçiler
@kingofzynx Smart take. Fitting into existing workflows another full replacement. Checking this out.
English
One thing i like here is that this isn't trying to replace the rest of your agent stack.
if it helps manage context while fitting into what people are already using, that's a much easier tool to evaluate.
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
@costeisnotbasic This is actually huge for debugging. No more black box frustration 😂
English
@fineboypsycho Same here, that 8% to 77% jump hit different. Local models are actually viable now.
English
seeing local models go from 8% to 77% tool accuracy at 100 tools definitely caught my attention. i wasn't expecting that number. gonna read through the benchmark and repo.
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
@pendingbilonia Agreed, that token drop sounds huge. Gonna benchmark it myself!
English
70–85% fewer input tokens at realistic tool catalog sizes is enough to make me curious
That's the kind of claim that's worth checking against the benchmark.
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
He really came prepared with a backup shirt the entire time. 💀
English
@olaoftrades Mind blown 80% token cuts without swapping models? This changes everything for agents. 🔥
English
cutting input tokens 70–85% without switching models shouldn’t be possible. benchmark is public on github.
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
@derxvyral This is huge for agent devs 7x cheaper without the hassle? Checking it out rn! 🚀
English
Ratel keeps agent context lean by only loading what’s required for the task. 7x cheaper according to their tests. Works on top of what you already run.
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
@matyCly Ratel dropping progressive disclosure like that? Game changer.
English
i like tools that solve boring problems. loading only the context an agent actually needs makes a lot more sense than stuffing everything into the prompt. going to read through the repo after work.
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
@zephyrarc0x Just starred the repo, this is huge for anyone building agents.
English
the local model numbers caught my attention more than the demo, 8% → 77% tool accuracy with 100 tools is worth reading the benchmark for.
cloning the repo tonight
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
@ThatF00dieGirl Just starred the repo, this is huge for anyone building agents.
English
cutting 70 to 85% of input tokens without changing models caught my attention.
i ended up reading the benchmark instead of scrolling.
worth checking the repo if you build agents.
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
repo is open access on github: github.com/ratel-ai/ratel
English
cutting input tokens by 70–85% at realistic tool catalog sizes without changing models. public benchmarks in repo
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English
repo is open access on github: github.com/ratel-ai/ratel
English
this would have been a $99/mo saas with a waitlist two years ago. open access repo on github now
Ratel@ratelai
We just made agents 7x cheaper to run Today we're releasing Ratel Right now your agent loads all the tools and skills you own to do one small task Ratel loads only what that task needs See how it works ↓
English