Abstract Euthanasia of laboratory animals is one of the main concerns for animal welfare, especially the use of carbon dioxide (CO 2 ) which has been described to induce aversive-like behaviours. In order to refine mass euthanasia in rodents, this study aims to determine which gas, sevoflurane, isoflurane or CO 2 induce less aversion in mice. Contrarily to carbon dioxide, the other gases do not induce formation of carbonic acid in the mucosa and are often used to anaesthetized mammals. A new approach-avoidance test was performed using 26 male NMRI mice. Animals had the possibility to choose between two gas environments delivered in different chambers at a 0.2L/min:1) 8% sevoflurane or 4% isoflurane (sevo vs iso), 2) 8% sevoflurane or 100% CO 2 (sevo vs CO 2 ) and 3) 4% isoflurane or 100% CO 2 (iso vs CO 2 ). Mice were placed in a central elevated compartment called the air chamber, and moved freely between the rewarded gas chambers until ataxia was observed or when 300s elapsed. Time spent in each chamber, the number of visits per minute, and eaten rewards were measured to evaluate aversion. Mice spent significantly more time in the sevoflurane chamber than in the air chamber in the sevo vs CO 2 treatment. In the iso vs CO 2 treatment, mice spent less time in the isoflurane chamber than in the air chamber. Also, the introduction of isoflurane and CO 2 in the testing trial induced mice to spend less time in these gas chambers compared with the time spent when no gas was present during the habituation period. Moreover, animals also ate fewer pellets in the isoflurane chamber compared with what was eaten when no gas was present. These observations indicate that sevoflurane may induce lower levels of aversion, contrarily to CO 2 , and isoflurane. Thereby the use of sevoflurane may contribute to the refinement of mice euthanasia. Sevoflurane may be used to induce unconsciousness before a potential more aversive gas is used to finish the euthanasia process.