Poor Paul's Benchmark
Open LLM inference benchmarks on real consumer hardware.
The raw data lives on Hugging Face. This site is the easiest way to get insights.
4773
Benchmark Results
88
Models
3
GPUs
1
Contributors
Leaderboard
See which model + GPU combos deliver the best throughput and lowest latency.
Explore
Interactive charts to visualize throughput, latency, and hardware comparisons.
Articles
Benchmark writeups, analysis posts, and hardware deep dives.
Want to contribute benchmarks from your own hardware?
Run the CLI →Data last updated: March 12, 2026 at 08:18 AM