True
The cost of running AI models — often called the "price of AI brains" — has dropped to about one-thirteenth of its level one year ago when measured at equal performance, according to data from Epoch AI.
The dramatic decline holds under a key premise: the comparison is made on a unit-cost basis for models of identical capability. Raw price cuts across differently capable models can be misleading, but when performance is held constant, the cost of obtaining a given level of AI output has collapsed to roughly a thirteenth of what it was a year earlier.
Epoch AI's analysis, presented in figures including "The Plunging Price of Thought," tracks the falling cost of AI inference over time. The data underpinning the one-thirteenth figure draws on multiple underlying data points and was reviewed against primary sources.
The claim stands only with the explicit condition of equal performance. Without that premise — comparing models of differing capability — the size of the price decline would not be directly comparable.
The claim that the price of AI inference fell to one-thirteenth in a year is True.