MLCommons proclaims outcomes from its MLPerf 4.0 inference benchmarks, which added Llama 2 70B and Stable Diffusion XL; programs with Nvidia GPUs got here out on high (Samuel Ok. Moore/IEEE Spectrum)