LLM inference costs have fallen by 9x to 900x per year depending on the task, with some fixed-performance LLM inference costs halving every two months.
Open the full topic