SimGPT is the first framework to embed Large Language Models directly inside a discrete event simulation engine (Simio), letting AI make decisions during a running model rather than only before or after it.
Traditional simulation treats AI as something you run around a model: tune parameters up front, or analyze results afterward. SimGPT removes that boundary by packaging a Large Language Model as a native Simio Step, so the engine can call the model mid run to dispatch jobs, interpret state, and generate data on the fly. It is provider agnostic, so the same Step works across model families and versions without rewriting the model.
The LLM is queried while the clock runs, so logic that once needed hard coded rules now reasons in context.
No external orchestration. Drop one file in and the SimGPT Step appears in the standard library.
Swap providers and versions without touching the simulation, with a path to self hosted open models.
An LLM driven dispatcher chooses the next job at each decision point and beats traditional scheduling heuristics.
SchedulingHow prompt design shapes the quality and stability of LLM decisions made inside a live run.
PromptingA controlled comparison across LLM families and versions and how model choice changes outcomes.
BenchmarkingGenerating emergency room inputs — arrivals, characteristics, length of stay — where real data is scarce.
Healthcare · three stepsPlace it into your Simio UserExtensions folder.
Open Simio — the SimGPT Step appears under User Defined Steps.