Oddly, I’ve lately been finding my small (35GB) Qwen3.6-35B-A3B-Q8 locally hosted model is very helpful under Pi Coding Agent doing PC admin, I’m kind of surprised.
And my big online AI (GLM-5.2) seems dumber than usual, so I reverted to Deepseek-V4-pro under Opencode. What’s going on ?
Here is an interesting article about SMALL AI and how it’s changing lives in small remote third world countries.
They don’t need, and don’t have the resources or the money for massive data centers.