Model benchmarks

pruna benchmarked several models to showcase the performance gains of the optimized versions.

P-Bench

The P-Bench leaderboard compares text-to-image, text-to-video, image-to-video, and video-to-video models on quality, speed, and price across inference providers.

Each provider may use their own optimization methods, and those details are not always public. Endpoints can also vary in latency, so benchmark them against your own use case.