Background and Purpose There is currently no medication approved specifically to treat cocaine addiction. Behavioural interventions such as cue exposure therapy (CET) rely heavily on new learning. Antagonism of the metabotropic glutamate 5 (mGlu5) receptor has emerged as a potential treatment, by reducing the reinforcing properties of cocaine. However, mGlu5 receptor activity is necessary for learning; therefore, such agents could interfere with behavioural treatments. We used a novel rodent model of CET to test the effects of mGlu5 negative and positive allosteric modulators (NAM and PAM) on behavioural therapy. Experimental Approach Rats were trained to press a lever for cocaine in the presence of a discrete cue [conditioned stimulus (CS)] and then extinguished in the absence of the CS. Following lever extinction, half the rats received CS extinction in the same chambers but with the levers withdrawn; the remaining rats received no CS extinction. Before this session, rats received a systemic administration of either vehicle or a mGlu5 NAM (MTEP, experiment 1) or PAM (CDPPB, experiment 2). Cue-induced reinstatement was tested in a drug-free session the following day. Key Results At reinstatement, rats that had received CS extinction showed reduced responding. This effect was attenuated by MTEP treatment before CS extinction. In contrast, administration of CDPPB (PAM) led to decreased reinstatement the following day, regardless of extinction condition. Conclusion and Implications These results suggest that mGlu5 receptor activity is both necessary and sufficient for efficient extinction of a cocaine-associated CS. Therefore, mGlu5 PAMs could enhance the efficacy of CET.