Local AI with a Mini PC is Totally Usable

Run local AI models on a mini PC with no dedicated GPU using LM Studio and the GMKtec M7's AMD 680M integrated graphics. In this video, I test multiple models including Qwen 3, Granite 4H Tiny, and OpenAI GPT-OSS getting 11-20 tokens per second — completely usable for coding, creative writing, and more. 0:00 - Intro 0:08 - GMKtec M7 & 680M Shared Memory Explained 0:40 - Task Manager GPU Memory Breakdown 1:37 - Loading Models in LM Studio 2:27 - Testing a 4B Model (Qwen 3) 2:54 - Coding Test: 20 Tokens/Second 3:21 - Loading a Larger Model (OpenAI GPT-OSS) 4:07 - Story Generation & Longer Prompts 5:05 - 128GB Mini PCs & OCuLink 6:11 - Thinking Model Test (Qwen 3 Thinking) 7:25 - Results & Comparison to 5060 Ti 7:48 - Getting Into AI on a Budget 8:46 - n8n & Future Plans 9:14 - Wrap Up Related Products (affiliate links help support my content at no extra cost to you, thanks!): - GMKtec M7 Mini PC: https://www.amazon.com/dp/B0DQKRXKS5?tag=hakehardware-20 - GMKtec EVO X2 (128GB): https://www.amazon.com/dp/B0F53MLYQ6?tag=hakehardware-20 - Samsung 990 EVO Plus 2TB SSD: https://www.amazon.com/dp/B0DHLBDSP7?tag=hakehardware-20 - Crucial 96GB DDR5 RAM: https://www.amazon.com/dp/B0C79K5VGZ?tag=hakehardware-20 Related Videos: - Testing Qwen3-coder-30b with a Mini-PC: https://youtu.be/y7IHTVa8lL8 - ComfyUI on AMD 780M with Zluda: https://youtu.be/LVjOoTJ4NiQ - AMD Radeon 680M Usable with ComfyUI: https://youtu.be/j5_yYmpNrNQ - Strix Halo + ROCm 7.1 + Ubuntu: https://youtu.be/vX7cYahvcPI - Ollama with AMD Support on Windows: https://youtu.be/AgLL-mmlG9Y #LocalAI #MiniPC #LMStudio