Qwen3.6 35B (2 min) vs Muse Glimmer 30B (4 min) on custom Llama.cpp build (RTX 5080)

Collecting trend data

The percentage is based on available heat signal, not comment count or independent people.