I would recommend to start withQwen 3.6 35B at maybe Q5, it should be fast in that setup. For intelligence Qwen 3.7 27b, is smarter but will run much more slow. Others also mention gemma 4, which might be worth a try.