Hardware Corner — Definitive GPU Ranking for LLMs (llama-bench CUDA 12.8, Q4_K_XL decode/PP vs context)
benchmark
Fields
id
sources/hardware-corner-gpu-ranking-local-llmPublisher
Hardware Corner
Kind
benchmark
Accessed
2026-10-14
notes
Hardware Corner in-lab llama-bench runs (Ubuntu 24.04, CUDA 12.8), Q4_K_XL quant. Decode tok/s per GPU per model at 16K/32K/65K/131K context. Third-party measured, single-GPU batch-1.
Notes
Hardware Corner in-lab llama-bench runs (Ubuntu 24.04, CUDA 12.8), Q4_K_XL quant. Decode tok/s per GPU per model at 16K/32K/65K/131K context. Third-party measured, single-GPU batch-1.
Cross-references
This record references no other records.