Per-query energy consumption of LLMs (via avsm) — discussion

#ai