diff --git a/vllm/model_executor/models/molmo.py b/vllm/model_executor/models/molmo.py
index feb585022317..b2154ef54af3 100644
--- a/vllm/model_executor/models/molmo.py
+++ b/vllm/model_executor/models/molmo.py
@@ -52,7 +52,8 @@
 from vllm.sequence import IntermediateTensors
 from vllm.utils import JSONTree, json_map_leaves
 
-from .interfaces import SupportsLoRA, SupportsMultiModal, SupportsPP
+from .interfaces import (SupportsLoRA, SupportsMultiModal, SupportsPP,
+                         SupportsQuant)
 from .utils import (AutoWeightsLoader, WeightsMapper, flatten_bn,
                     is_pp_missing_parameter,
                     make_empty_intermediate_tensors_factory, make_layers,
@@ -633,7 +634,8 @@ def forward(
         return hidden_states, residual
 
 
-class MolmoVisionBackbone(nn.Module):
+class MolmoVisionBackbone(nn.Module, SupportsQuant):
+    packed_modules_mapping = {"merged_linear": ["gate_proj", "up_proj"]}
 
     def __init__(
         self,
@@ -794,7 +796,7 @@ def load_weights(self, weights: Iterable[Tuple[str,
 
 
 @support_torch_compile
-class MolmoModel(nn.Module):
+class MolmoModel(nn.Module, SupportsQuant):
 
     def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
         super().__init__()
@@ -1402,8 +1404,8 @@ def get_replacement_molmo(item_idx: int):
 @MULTIMODAL_REGISTRY.register_processor(MolmoMultiModalProcessor,
                                         info=MolmoProcessingInfo,
                                         dummy_inputs=MolmoDummyInputsBuilder)
-class MolmoForCausalLM(nn.Module, SupportsMultiModal, SupportsPP,
-                       SupportsLoRA):
+class MolmoForCausalLM(nn.Module, SupportsMultiModal, SupportsPP, SupportsLoRA,
+                       SupportsQuant):
     hf_to_vllm_mapper = WeightsMapper(
         orig_to_new_substr={
             # vision backbone mapping