Run Qwen 3.5/3.6 35B on 8GB VRAM | LM Studio + Opencode Setup (40 tk /s)

Описание: UPDATE: This will also work with the new Qwen 3.6 models!

Waffle ends at 1:35, 1.5x speed recommended.
Run large models on cheap GPUs. This model is on-par with Claude Haiku 4.5, and can run on cheap consumer hardware!
Setup guide for LM Studio + Opencode. You can also use ollama.
https://qwen.ai/blog?id=qwen3.5