Gartner projects open-weight models will run 90 percent of enterprise agent steps by 2030

Gartner analysts Mark Petty and Raymond Paquet wrote that typical operational workflows contain hundreds of routine steps such as classification, tool calling, and data validation. They state open-weight models handle these at parity for a quarter the price of the cheapest closed model and as little as a twentieth the price of a frontier model run on every step.

The analysts note current open models can be more verbose and retry more often, reducing the realized saving per task to roughly five or six times against a frontier model. They still describe the cost reduction as a welcome byproduct rather than the primary reason to adopt open weights.

Gartner offers the planning assumption that by 2030 ninety percent of enterprise operational agent workflow steps will execute on open-weight models. The report emphasizes that years of workflow history constitute valuable data that should remain inside the enterprise perimeter.

The analysts position frontier models as capital expense for research and development while open-weight models become operational expense for routine execution. Harnesses that route work to the right model are described as the main source of product value.