I built a local GPU benchmark server for 4B to 35B models because I was tired of synthetic cloud leaderboards and tiny sample tests · AI Intel