Decision Models (DMs) are a new class of AI systems that return a probability distribution over a fixed set of options with very low latency. Because of their speed and natural output format, they are proposed as human surrogates in economic simulations and surveys, a role that currently relies on large language models (LLMs) which are expensive at scale and often miss human behavioural variation. This work provides the first systematic evaluation of DMs as economic surrogates, comparing ten DMs with eight LLMs across three settings: a household consumption spending survey, several classic economic games, and a labour‑market field experiment. In every setting the top‑performing “frontier” DMs—those scoring highest on general decision‑making benchmarks—reproduced human response variation more faithfully than LLMs. They matched the modal human response in the spending survey, though they lagged slightly behind LLMs in the game scenarios. Frontier DMs were roughly ten times faster than the fastest LLM and could be over a hundred times cheaper. Together these findings suggest that DMs capture human behavioural heterogeneity better than LLMs and make large‑scale economic simulations far more feasible.\n\nReview