AnalysisDevelopersNovember 22, 2026

llama.cpp adds support for Nemotron 3.6 Academia

Read original source →reddit.com

A llama.cpp PR adds support for Nemotron 3.6 Academia, a looped hybrid model combining Gated Delta Product, dense attention, and MLP layers. It also supports a last_loop KV cache, as with Dory.

1 source

More stories today

Open the live feed