How I got Qwen 3.8 Flash Next running at 10 tokens/sec on an RTX 5060 Ti, and why paging n-gram tables from NVMe avoids the brutal disk bandwidth bottlenecks of MoE expert streaming.
How to securely publish local LM Studio models to the internet using a Dockerized LiteLLM proxy and Cloudflare Tunnels.
A visual deep-dive into training a custom transformer LLM from scratch with PyTorch, analyzing loss curves, overfitting, and hyperparameter tuning.
A hands-on setup guide for running Clawbot against local LM Studio models and securely exposing the web interface using Cloudflare Tunnels.
Colibrì vs. FreeToken: Can Consumer Hardware Really Run Frontier MoE Models?
Does Windows Dev Drive Actually Speed Up Local LLMs? A Deep Dive into ReFS, Defender, and NVMe Paging
Getting 10 tok/s on Qwen 3.8 Flash Next: Why N-Gram Disk Paging Works Where MoE Fails
I love to hear from you!
GaⅤin.com by Gavin W.