weddle/Qwen3.8-27B-AutoRound-W4A16-G128

Clarify decode timing in headline caption

Lead with quantization purpose and single-B65 performance highlights

Add measured short and long context prefill rates

Document single-B65 throughput hardware and context capacity

Improve model card layout and explain packed tensor display

Add AutoRound W4A16 G128 artifact and model card

initial commit