diff --git a/Dockerfile b/Dockerfile
index df79412bbeceb..21945cb0bd09c 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -286,7 +286,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
     if [ "$TARGETPLATFORM" = "linux/arm64" ]; then \
         uv pip install accelerate hf_transfer 'modelscope!=1.15.0' 'bitsandbytes>=0.42.0' 'timm==0.9.10' boto3 runai-model-streamer runai-model-streamer[s3]; \
     else \
-        uv pip install accelerate hf_transfer 'modelscope!=1.15.0' 'bitsandbytes>=0.45.0' 'timm==0.9.10' boto3 runai-model-streamer runai-model-streamer[s3]; \
+        uv pip install accelerate hf_transfer 'modelscope!=1.15.0' 'bitsandbytes>=0.45.3' 'timm==0.9.10' boto3 runai-model-streamer runai-model-streamer[s3]; \
     fi
 
 ENV VLLM_USAGE_SOURCE production-docker-image
diff --git a/docs/source/features/quantization/bnb.md b/docs/source/features/quantization/bnb.md
index b81d89c457513..fc499e7692d98 100644
--- a/docs/source/features/quantization/bnb.md
+++ b/docs/source/features/quantization/bnb.md
@@ -9,7 +9,7 @@ Compared to other quantization methods, BitsAndBytes eliminates the need for cal
 Below are the steps to utilize BitsAndBytes with vLLM.
 
 ```console
-pip install bitsandbytes>=0.45.0
+pip install bitsandbytes>=0.45.3
 ```
 
 vLLM reads the model's config file and supports both in-flight quantization and pre-quantized checkpoint.
diff --git a/vllm/model_executor/layers/quantization/bitsandbytes.py b/vllm/model_executor/layers/quantization/bitsandbytes.py
index 33c2ca93ffa17..1e8e7aa1b8c12 100644
--- a/vllm/model_executor/layers/quantization/bitsandbytes.py
+++ b/vllm/model_executor/layers/quantization/bitsandbytes.py
@@ -155,12 +155,12 @@ class BitsAndBytesLinearMethod(LinearMethodBase):
     def __init__(self, quant_config: BitsAndBytesConfig):
         try:
             import bitsandbytes
-            if bitsandbytes.__version__ < "0.45.0":
+            if bitsandbytes.__version__ < "0.45.3":
                 raise ImportError("bitsandbytes version is wrong. Please "
-                                  "install bitsandbytes>=0.45.0.")
+                                  "install bitsandbytes>=0.45.3.")
         except ImportError as err:
-            raise ImportError("Please install bitsandbytes>=0.45.0 via "
-                              "`pip install bitsandbytes>=0.45.0` to use "
+            raise ImportError("Please install bitsandbytes>=0.45.3 via "
+                              "`pip install bitsandbytes>=0.45.3` to use "
                               "bitsandbytes quantizer.") from err
 
         self.quant_config = quant_config
diff --git a/vllm/model_executor/model_loader/loader.py b/vllm/model_executor/model_loader/loader.py
index b2ffca2a4b4dc..d3f7a26e7f9e0 100644
--- a/vllm/model_executor/model_loader/loader.py
+++ b/vllm/model_executor/model_loader/loader.py
@@ -862,12 +862,12 @@ class BitsAndBytesModelLoader(BaseModelLoader):
         try:
             import bitsandbytes
 
-            if bitsandbytes.__version__ < "0.45.0":
+            if bitsandbytes.__version__ < "0.45.3":
                 raise ImportError("bitsandbytes version is wrong. Please "
-                                  "install bitsandbytes>=0.45.0.")
+                                  "install bitsandbytes>=0.45.3.")
         except ImportError as err:
-            raise ImportError("Please install bitsandbytes>=0.45.0 via "
-                              "`pip install bitsandbytes>=0.45.0` to use "
+            raise ImportError("Please install bitsandbytes>=0.45.3 via "
+                              "`pip install bitsandbytes>=0.45.3` to use "
                               "bitsandbytes quantizer.") from err
 
         hf_weights_files, use_safetensors = self._prepare_weights(