Genome-wide array-based analyses of tobacco use have identified many associated common variants, but the vast majority of the heritability remains unexplained. Rare variants likely contribute to risk for smoking and other addictive behaviors, and cost-effective whole genome sequencing is now making genome-wide analyses of rare variants possible. The present study reports rare variant association results for tobacco use from TOPMed, which has at present called genotypes from 9,536 30x whole genome sequences in 8 distinct studies with smoking phenotypes. We conducted a genome-wide association study (GWAS) of all discovered variants with minor allele count > 10 for four phenotypes: regular vs never smoker (n=9,536;n variants=34,115,958), current vs former smoker (n=4,601;n variants=27,243,212), age of initiation of regular smoking (n=4,467;n variants=27,141,662) and cigarettes smoked per day (n=4,590;n variants=27,547,744). The GWAS yielded no significant results after Bonferroni correction per phenotype, although there was a signal suggested for the known variant rs16969968 in CHRNA5 (p=7.718e-05). Rare variant burden tests of highly deleterious variants also yielded no significant results. Finally, we used the TOPMed sequences to fine map 182 recently discovered smoking-related loci--defined as 1MB regions around top associated variants--from array-based GWAS meta-analyses across all four smoking phenotypes (N ranging from 256,658 to 991,257 depending on the phenotype). Of the 182 most significantly associated common variants within these 182 loci, two were significant after Bonferroni correction for 182 tests (rs56113850 in CYP2A6, p=7.3e-5 and rs112878080 in CHRNA3, p=2.6e-5), and 121 of the 182 variants showed the same direction of effect in TOPMed and GSCAN (p=5.1e-6). Forward selection fine mapping tests in TOPMed within the 182 loci discovered one additional conditionally independent variant in an intron of ASXL1 (rs148986225, MAF=0.003, p=1.7e-7) associated with cigarettes smoked per day after Bonferroni correction per phenotype. We expect to present updated results on a much larger data freeze approaching 50,000 total individuals, greatly increasing power and precision of all analyses.