Background: Cannabis is increasingly legal and available to cancer patients, but the pattern of cannabis use among cancer patients is unknown. We aimed to characterize cannabis use patterns and beliefs among subgroups of cancer patients. Methods: We identified adults who received treatment between March and August 2021 for any of 10 cancers (brain, breast, head and neck, gynecological, gastrointestinal, lymphoma, prostate, testicular, and lung) at Memorial Sloan Kettering Cancer Center. We included patients residing in New York, New Jersey, or Connecticut, where medical and recreational cannabis use is legal. We contacted patients by email, phone, and mail to invite them to complete an anonymous online or phone survey regarding experiences with cannabis. We administered surveys from August 2021 to April 2022. The survey used harmonized measures developed by investigators from 12 NCI-designated cancer centers to elicit patient-reported cannabis use, Cannabis was defined as any cannabis product, including CBD-only products. Weighting was assigned based on characteristics (age, sex assigned at birth, race, ethnicity, cancer) of the sampled hospital population to account for potential nonresponse bias. We used weighted multivariable logistic regression to identify self-reported sociodemographic predictors (gender, age, race, ethnicity, income, insurance, marital status, employment status) and clinical predictors (cancer) of cannabis use during treatment. Results: 1258 patients completed the survey (35% response rate). 75% of respondents were white non-Hispanic; 51% were under age 65. Regardless of cannabis use during treatment, 82% (N = 1032) perceived benefits of cannabis use. The most commonly endorsed benefits were relief of stress, anxiety, or depression (68%), pain management (66%), sleep improvement (48%), and managing cancer side effects (42%). 53% of patients (N = 666) perceived risks to cannabis use; commonly endorsed risks were inability to drive (24%), difficulty concentrating (23%), addiction (22%), and lung damage (21%). Across all respondents, 22% (N = 278) reported cannabis use during treatment, and 53% used cannabis a few times a week or more frequently. Weighted multivariable models (N = 991) showed that accounting for all predictors, only age showed a significant association with cannabis use. Patients ≤45 years old were more likely to use cannabis than patients ≥65 years old (odds ratio 4.65, 95% confidence interval 2.11-10.23). Conclusions: In a large sample of cancer patients, a large proportion reported beliefs that cannabis can benefit symptoms relevant to cancer, and about half of patients acknowledged potential harms. Younger patients were significantly more likely to use cannabis during treatment than their older counterparts. This may represent a generational shift in attitudes toward cannabis, as access to medical and recreational cannabis expands.