Various subtypes of metabotropic glutamate receptors (mGluRs) have been implicated in fear extinction, but mGluR2/3 subtype has not been tested. Here, we found that microinjection of an mGluR2/3 antagonist, LY341495, into the lateral amygdala (LA), but not into the adjacent central amygdala (CeA), impaired extinction retention without affecting within-session extinction. In contrast, we failed to detect any significant changes in motility and anxiety during a period when extinction training or retention was performed after LY341495 injection, suggesting that the effect of LY341495 is specific to conditioned responses. Subsequently, on the basis of a previous finding that a long-term potentiation of presynaptic efficacy at cortical input synapses onto the lateral amygdala (C-LA synapses) supports conditioned fear, we tested the hypothesis that activation of mGluR2/3 leads to fear extinction via a long-term weakening of presynaptic functions at C-LA synapses. Fear extinction produced a decrease in C-LA synaptic efficacy, whereas LY341495 infusion into the LA blocked this extinction-induced C-LA efficacy decrease without altering synaptic efficacy at other LA synapses. Furthermore, extinction enhanced paired pulse ratio (PPR) of EPSCs, which inversely correlates with presynaptic release probability, whereas LY341495 infusion into the LA attenuated the extinction-induced increase in PPR, suggesting the presence of mGluR2/3-dependent presynaptic changes after extinction. Consistently, extinction occluded a presynaptic form of depression at C-LA synapses, whereas the LY341495 infusion into the LA rescued this occlusion. Together, our findings suggest that mGluR2/3 is required for extinction retention and that the mGluR2/3 action is mediated by the long-term weakening of release probability at C-LA synapses.