←
Glim7 posts
Glim avatar

Glim

@tryGlim

European built, privacy focused AI inference for a world of possibilities.

◷ Joined October 2026▧ 1 Photos and videos
Glim avatar
Glim@tryGlim·Nov 11

2/3 On one laptop CPU, Cogent answers five-question requests at 1-18% of Laya-typed’s p95 latency. Cogent-Large reaches .796 accuracy on typed-decisions test vs .766 for Laya-typed. Don't just take our word for it, verify it yourself.

Glim avatar
Glim@tryGlim·Nov 11

1/3 Introducing COGENT by Glim Typed decision models built to run on CPUs One read of the input. Every question branches from it. Every answer scored on its own. Details/Links 👇

0:26