diff --git a/vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe/compressed_tensors_moe_wna16_marlin.py b/vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe/compressed_tensors_moe_wna16_marlin.py
index 0401a5b6e73..46fa36180d9 100644
--- a/vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe/compressed_tensors_moe_wna16_marlin.py
+++ b/vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe/compressed_tensors_moe_wna16_marlin.py
@@ -80,7 +80,6 @@ class CompressedTensorsWNA16MarlinMoEMethod(CompressedTensorsMoEMethod):
             else:
                 scale = kInt4StaticGroupScale
         elif self.num_bits == 8:
-            assert self.group_size == -1
             scale = kInt8StaticGroupScale
         else:
             raise ValueError(