Benchmark of open LLM inference cost on consumer hardware (CUDA/Apple Silicon/CPU) using Hugging Face Transformers - View it on GitHub