paulplee/poor-pauls-benchmark
Run a standardized benchmark of GPU performance against GGUF LLMs (throughput, TTFT/TTI-like latencies, ITL, and VRAM limits) across quantizations and context sizes, and submit results to a public leaderboard.
Python (likely, based on typical benchmark tooling for GGUF ecosystems)GGUF ecosystem tooling (e.g., llama.cpp-compatible runtime)GPU compute stack (CUDA/ROCm depending on environment)2mo ago
brand newby paulpleeFR:MEDPDR:HIGHMCR:HIGHDH:6MO2/10