Been playing with LLMs lately on my own self-hosted Ollama instance. Very surprised to see that Intel 12th gen iGPUs can run small models relatively well. It takes time to load, but it's quite usable once loaded. #llm #ollama #intel #igpu