Poor Paul's Benchmark: leaderboard

Paul Lee's local-inference benchmark for GGUF models on consumer GPUs, sweeping quantization, context length and concurrency to log throughput, TTFT, inter-token latency and VRAM limits.

Source: github.com.

Interactive version: theaggregate.ai/benchmark?slug=poor-paul-s-benchmark · How It Works · Data refreshed daily, snapshot 2026-09-05.