llama.cpp fails to load Unsloth Qwen3.8 after PC restart

Mashic · reddit · 2026-08-15

A user reports a bug where llama.cpp on Windows fails to load the Unsloth Qwen3.8-27B-GGUF model after a PC restart. Using the Hugging Face alias triggers a redownload, while the local path fails entirely. Other models function correctly.

Original post →

More from Infra

Infra channel →