From Words to Watts: Benchmarking the Energy Costs of LLM Inference (2023) · HackerLangs