Weekend project success!
✅ run Qwen3.5 35B on 3090 on old machine via llama.cpp
✅ use it from laptop on local network
Not only it works but it is very capable and it gives me over 100 tokens / s ( 2000+ for prompt eval)
It’s just… awesome.
Thanks
@gladstein and
@Justin Moon for the inspiration from the podcast on sovereign AI.
Next steps: use it securely from anywhere , find more old “cheap” hardware and set up more models, use them from Hermes agent … onward