Local AI

Thirty billion on the desk, three billion awake

Ollama shipped NVIDIA Nemotron 3.5 Lightning, a 30B open model that keeps about 3B parameters active. That is a local agent you can actually budget hardware for.