Qwen 3.5 Setup on Your Local Computer (Step-by-Step Guide)

Summary

This video walks you through setting up Qwen 3.5 on your local computer using LM Studio, a tool the hosts prefer over Ollama for its model browser and ease of use. You start by downloading LM Studio, which works on both Windows and Mac, and installing it with a straightforward next-through setup process. Once installed, you open the model browser and search for Qwen 3.5. Depending on your hardware, you choose a model size that fits your available GPU or RAM — the hosts demonstrate using the 9 billion parameter version (around 6 GB) on an Nvidia RTX 3060 for full GPU offload and faster inference. On Apple hardware, unified memory handles both the model and the system, so more RAM directly translates to larger, better-performing models. After the model downloads, you load it into a new chat session inside LM Studio and start querying it directly. The hosts benchmark the setup at around 37 tokens per second on the 3060, which they note is reasonably fast for local inference. They also briefly show how you can connect the local model to OpenClaw once LM Studio is running, though that integration is saved for a follow-up video. The video is honest about trade-offs: local models give you full data privacy since nothing leaves your machine, and they are free to run, but they do not yet match the raw intelligence of top-tier cloud models like Claude or GPT-4. The hosts test Qwen 3.5 with a simple logic question and find the model over-thinks it, taking over two minutes before crashing — a reminder that smaller local models still have real limitations. The key takeaway is that if privacy matters to you, or you want a free, offline AI assistant for basic tasks, LM Studio plus Qwen 3.5 is a viable and accessible setup for most modern PCs and Macs.