Did AI inference costs fall or rise since 2022?

Since 2022, AI inference costs have fallen^[1]. Between 2022 and 2024, the cost-per-token to run language models fell by an estimated 99.7%^[1]. This decline was driven by improvements in both hardware and algorithmic efficiency^[1].

As inference becomes cheaper and more efficient, the competitive pressure amongst LLM providers increases^[1]. What used to cost dollars can now cost pennies, and what cost pennies may soon cost fractions of a cent^[1].