Launched in 2004, the Autism Genome Project is the largest study ever conducted to find the genes associated with inherited risk for autism. The project is a public/private research partnership involving approximately 50 academic and research institutions that have pooled their DNA samples in a collaborative effort. An initial AGP research project, Phase 1, was completed in 2007 and included linkage analysis in a sample of over 1,400 multiplex families - each with two or more individuals affected by autism spectrum disorders (ASDs). Extensive phenotypic information was gathered on the individuals in each family, creating the most comprehensive database of autism families at the time. The second AGP project, Phase 2, is ongoing and will collect detailed information on a new sample of over 3,000 families, in which at least one child has an ASD. As a result of this research investigators have identified multiple rare mutations and estimated that approximately 15 per cent of individuals with an ASD have an identifiable genetic etiology. It is now known that a substantial proportion of genetic risk for ASDs resides in rare variants associated with high odds ratios for risk. These rare variants include variation across a broad range, from single base changes (point mutations), indels and microindels, CNVs, and larger chromosomal imbalances. The finding that rare variants have a role in ASDs represents a major step forward in understanding of the genetic architecture of ASDs. In the next phase (Phase 3) the AGP proposes a comprehensive set of studies that will lead to the discovery of new variations and the translation of the findings into clinical practice. The first objective of the Phase 3 Program will be to extend recently completed studies in 3,000 well-characterized ASD trios (Phase 2) and ongoing high throughput sequencing of 1,200 samples from within this sample to provide large-scale confirmation of the autism risk factors identified by the AGP and to inform clinical practice guidelines. The second objective will be to establish a comprehensive list of genetic variants predisposing to ASDs. The only way to fully characterize the entire spectrum of variation in the human genome is by sequencing large numbers of individuals. This work will inform development of more comprehensive and available genome arrays for autism diagnostic approaches in the future.