Fentanyl overdose deaths have reached “epidemic” levels in North America. Death in opioid overdose invariably results from respiratory depression. In the present work, we have characterized how fentanyl depresses respiration, and by comparing fentanyl with heroin and morphine, the active breakdown product of heroin, we have sought to determine the factors, in addition to high potency, that contribute to the lethality of fentanyl. Respiration (rate and tidal volume) was measured in awake, freely moving mice by whole body plethysmography. Intravenously administered fentanyl produced more rapid depression of respiration than equipotent doses of heroin or morphine. Fentanyl depressed both respiratory rate and tidal volume. Fentanyl did not depress respiration in μ‐opioid receptor knockout mice. Naloxone, the opioid antagonist widely used to treat opioid overdose, reversed the depression of respiration by morphine more readily than that by fentanyl, whereas diprenorphine, a more lipophilic antagonist, was equipotent in reversing fentanyl and morphine depression of respiration. Prolonged treatment with morphine induced tolerance to respiratory depression, but the degree of cross tolerance to fentanyl was less than the tolerance to morphine itself. We propose that several factors (potency, rate of onset, lowered sensitivity to naloxone, and lowered cross tolerance to heroin) combine to make fentanyl more likely to cause opioid overdose deaths than other commonly abused opioids. Lipophilic antagonists such as diprenorphine may be better antidotes than naloxone to treat fentanyl overdose.