Tokenmaxxing is so over. It's all about modelmaxxing now.

Tokenmaxxing involved encouraging employees to use AI as much as possible, while modelmaxxing directs engineers to use specific AI models for specific tasks to improve efficiency. The shift follows acknowledgments from some founders that the earlier approach led to wasted tokens.

The new strategy aims to reduce costs and improve performance by routing queries to the most appropriate model rather than always using a large, general-purpose model. The report notes that this transition reflects a broader maturation of AI deployment within enterprises, as companies seek to balance productivity gains with cost control.

The exact timeline of the shift was not specified, but it appears to be a recent development as of mid-2026. The article highlights that the change is driven by the need for more efficient use of AI resources, rather than maximizing volume. The move from tokenmaxxing to modelmaxxing is expected to influence how companies train employees and allocate AI budgets.