gpt-oss:120b

A hobbyist running local large language models on a Linux box with a Blackwell GPU (96GB VRAM) and 125GB CPU RAM reports that gpt-oss:120b is their preferred coding model, though Qwen3.8:27b recently …

A new llama.cpp patch called HotPin enables running a 120-billion-parameter mixture-of-experts (MoE) model on just 24GB of RAM, achieving up to 67% memory savings and a 45% speedup over standard swapp…