Smoking is a major preventable cause of morbidity and mortality in the United States and is moderately heritable. Genome wide association studies (GWAS) has allowed the discovery of hundreds of common variants associated with smoking. Although low frequency variants are expected to influence tobacco use, these rare variants are yet to be identified possibly due in part to high statistical power required to test rare variants. Most of the GWAS for smoking have been conducted in individuals with European ancestry and similar associations have not been studied in non-European populations. As a part of the Trans-Omics for Precision Medicine Whole Genome Sequencing Program (TOPMed), we seek to identify rare variants associated with four measures of smoking behaviors in a multi-ethnic sample. Among the four phenotypes, two relate to the initiation of smoking, capturing the age at which they started smoking regularly and whether an individual has ever become a regular smoker in their life. The remaining two relate to smoking cessation which contrasts current and former smokers. Finally, the number of cigarettes smoked per day reflects the heaviness of tobacco use among smokers. At the time of writing, the TOPMed project has called genotypes from 65,000 individuals with 30x whole genome sequences. A subset of 45,000 individuals from this data freeze have one or more smoking phenotypes. We are conducting single variant association tests for age of initiation, smoking initiation, smoking cessation, and cigarettes smoked per day with 154,176,919 variants with minor allele count >=1 using whole genome sequences from 7,067 individuals. The preliminary results for age of initiation showed no significant associations, suggesting that additional samples will be required to raise the statistical power to make discoveries for this and other complex substance use phenotypes. We will present updated results from single variant tests and gene based tests on larger samples, up to the full sample of 45,000 for one or more phenotypes. Stronger power to detect effects will be achieved as the remainder of our sample is analyzed and as the data available to us continues to grow with the release of new data freezes through TOPMed.