diff --git a/vllm/model_executor/model_loader.py b/vllm/model_executor/model_loader.py
index 951ba1f0..82c8dfa7 100644
--- a/vllm/model_executor/model_loader.py
+++ b/vllm/model_executor/model_loader.py
@@ -35,6 +35,7 @@ _MODEL_REGISTRY = {
 # FIXME(woosuk): Remove this once all models support quantization.
 _MODEL_CLASSES_SUPPORT_QUANTIZATION = [
     LlamaForCausalLM,
+    MistralForCausalLM,
 ]