Hardware Corner: RTX 6000 Ada, llama.cpp CUDA, Qwen3 8B Q4_K XL GGUF, prompt-processing 4270.99 t/s @16k ctx

THIRD-PARTY MEASURED (hardware-corner.net local-LLM GPU ranking, accessed 2026-10-14). llama.cpp CUDA build, Qwen3 weights in 'Q4K XL' GGUF (their standard quant; nearest repo format gguf-q4-k-m), batch 1. Metrics are per-request decode tok/s (tg, output tokens/s for a single stream) and prompt-processing pp512 tok/s; both measured at ~16384-token context. NOT aggregate batch throughput. pp512 4270.99 t/s at 16k context for RTX 6000 Ada.

Prefill-side measure. pp512 on consumer cards scales with compute; 4090 tops this list at ~5.2k t/s.