← Glossary Company
PrismML
A startup that shrinks large open AI models small enough to run on a phone.
PrismML works on compression: taking a model built for server hardware and squeezing it until it fits on a consumer device without losing much of what it could do. Its Bonsai 27B release packed a 27-billion-parameter open model into under four gigabytes while keeping roughly 90 percent of its ability.
The appeal of that trick is privacy and independence. A model running on your own phone needs no internet connection and sends nothing to anyone’s servers.
Mentioned in