03.1Edge language models
Small models, real work
Compact language models tuned to run fast on phones, laptops and private servers, without sending everything to the cloud.
01Now building
Foundational AI
for the edge.
Pebblepin builds compact language models and retrieval systems that run on your device or close to it, and ships them as apps and software people use every day.
03What we build
Models, retrieval, products.
One stack, from the model up to the app in your hand.
03.2Retrieval systems
Find the source first
Search and retrieval that finds the right material before answering, so every answer can show where it came from.
03.3Apps and software
Built on our own stack
Mobile, desktop and web products powered by our models and retrieval, designed for everyday use.
04Approach
How we design.
APrivate by default
Keep data close
We design for processing on the device or inside your own walls wherever we can.
BGrounded answers
Show the source
We build retrieval first, so answers point back to the material they came from.
CFast and frugal
Small beats slow
Smaller models mean lower cost and instant responses, on hardware people already own.
05Contact
Our first products
are on the way.
Want early access, or to work with us? Send a note and we'll get back to you.
hello@pebblepin.com