The Collective Hub of Engineers and Internet Innovators
Twitter GitHub Member login
EngineerIsaac logo EngineerIsaac

Interactive Lab

Weather Radar timeline Rack Builder Interactive rack layout Tab Explorer Windows file explorer with tabs Aecii Image Browser ComfyUI image browser with remote updates LLM Core Builder Project skeleton for any model AI Constructs Open-weight models with direct download

Interactive Lab

AI Constructs

Open-weight models you can actually take home. Each card says what the model is for, whether a graphics card is required, the minimum that will run it, and the card that runs it well. Chat GGUFs can crawl on CPU. Image models need a real GPU. Numbers assume Q4_K_M with a normal context — longer context uses more VRAM.

0 constructs Direct Hugging Face downloads Not hosted on this server

Hardware notes assume Q4_K_M and a normal context window. More context, vision, and Q5/Q8 all need extra VRAM. Chat GGUFs can run on CPU through llama.cpp or LM Studio; that is usually too slow to enjoy above ~9B. Image models (FLUX, Qwen Image Edit) need a dedicated NVIDIA or Apple GPU. AMD works in some runtimes, but NVIDIA 12–24GB is the common path. “Minimum that will run” means it will load and answer. “Runs best on” means full GPU offload at a usable speed. Licenses still apply — Apache and MIT are the easy ones, Llama and some Kimi/Qwen variants have extra terms.