diff --git a/vllm/model_executor/layers/fused_moe/experts/deep_gemm_moe.py b/vllm/model_executor/layers/fused_moe/experts/deep_gemm_moe.py
index 3b354dd3ef13..4662bd690af3 100644
--- a/vllm/model_executor/layers/fused_moe/experts/deep_gemm_moe.py
+++ b/vllm/model_executor/layers/fused_moe/experts/deep_gemm_moe.py
@@ -355,9 +355,9 @@ def activation_format() -> mk.FusedMoEActivationFormat:
     def _supports_current_device() -> bool:
         from vllm.platforms import current_platform
 
-        return (
-            is_deep_gemm_supported()
-            and current_platform.is_device_capability_family(100)
+        return is_deep_gemm_supported() and (
+            current_platform.is_device_capability_family(100)
+            or current_platform.is_device_capability_family(120)
         )
 
     @staticmethod
diff --git a/vllm/platforms/cuda.py b/vllm/platforms/cuda.py
index 57814d29bef9..bfd84171c8c2 100644
--- a/vllm/platforms/cuda.py
+++ b/vllm/platforms/cuda.py
@@ -555,8 +555,11 @@ def support_static_graph_mode(cls) -> bool:
 
     @classmethod
     def support_deep_gemm(cls) -> bool:
-        """Currently, only Hopper and Blackwell GPUs are supported."""
-        return cls.is_device_capability(90) or cls.is_device_capability_family(100)
+        """Currently, Hopper, datacenter Blackwell (SM 100+) and consumer
+        Blackwell (SM 12x — RTX 50-series, GB10/DGX Spark) are supported."""
+        return (cls.is_device_capability(90)
+                or cls.is_device_capability_family(100)
+                or cls.is_device_capability_family(120))
 
     @classmethod
     def is_integrated_gpu(cls, device_id: int = 0) -> bool: