OBJECTIVES: Every winter season an influenza epidemic occurs, although strength and duration may vary. In 2006-2007 in Austria presumably 5% of the whole population fell sick while 21% of age 15 and above were vaccinated. The goal was to build an agent based model to understand, model and simulate the progress of influenza epidemics. METHODS: The agent based model simulates single persons with an infection state (susceptible, infected with or without symptoms, resistant, vaccinated). Based on the results of a wide European study (POLYMOD, EC-Project SP22-CT-2004-502084) people have contacts in different places like housholds, schools or workplaces. Transmissions are possible upon contacts, then a person is infected for a while until he or she becomes resistant upon recovery. RESULTS: The outbreak of the epidemic starts when a few people are initially infected while the rest is susceptible or vaccinated. After some time the epidemic stops due to a larger number of resistant and a smaller number of susceptible people. Since only 5% of the population fall sick the situations at outbreak and at termination of the epidemic are similar and therefore it behaves very sensitive to parameter changes. CONCLUSIONS: Some parameter changes in the model can be interpreted as interventions in reality. But usually the influenza does not react sensitive to interventions. For example, an increase of the vaccination rate by 5% prevents an outbreak of the epidemic in the model which is obviously not true. This insight has two consequences: First, the influenza does not just spread and stop by transmission and recovery of people. There must be one or more other impacts modulating outbreaks like predestined people to fall sick or the climate. Second, without knowledge of these impacts it is almost impossible to predict the effect of vaccination strategies exactely.