diff --git a/.github/workflows/format_pr_body.yaml b/.github/workflows/format_pr_body.yaml
index a99d812..24c087d 100644
--- a/.github/workflows/format_pr_body.yaml
+++ b/.github/workflows/format_pr_body.yaml
@@ -36,7 +36,7 @@ jobs:
 
       - name: Get vLLM version
         run: |
-          VLLM_COMMIT=17b4c6685ce62d5652654784d6771a3d38e4273e
+          VLLM_COMMIT=releases/v0.11.0
           echo "VLLM_COMMIT=https://github.com/vllm-project/vllm/commit/$VLLM_COMMIT" >> $GITHUB_ENV
 
       - name: Checkout repository
diff --git a/.github/workflows/vllm_ascend_test.yaml b/.github/workflows/vllm_ascend_test.yaml
index a45d676..7470694 100644
--- a/.github/workflows/vllm_ascend_test.yaml
+++ b/.github/workflows/vllm_ascend_test.yaml
@@ -42,7 +42,7 @@ jobs:
   lint:
     uses: ./.github/workflows/pre-commit.yml
     with:
-      vllm: 17b4c6685ce62d5652654784d6771a3d38e4273e
+      vllm: releases/v0.11.0
 
   changes:
     runs-on: ubuntu-latest
@@ -83,7 +83,7 @@ jobs:
         VLLM_USE_MODELSCOPE: True
     strategy:
       matrix:
-        vllm_version: [17b4c6685ce62d5652654784d6771a3d38e4273e, v0.10.2]
+        vllm_version: [releases/v0.11.0, v0.10.2]
     steps:
       - name: Install packages
         run: |
@@ -138,7 +138,7 @@ jobs:
     name: e2e-light
     strategy:
       matrix:
-        vllm_version: [17b4c6685ce62d5652654784d6771a3d38e4273e, v0.10.2]
+        vllm_version: [releases/v0.11.0, v0.10.2]
     # Note (yikun): If CI resource are limited we can split job into two chain jobs
     needs: [lint, changes]
     # only trigger e2e test after lint passed and the change is e2e related with pull request.
diff --git a/.github/workflows/vllm_ascend_test_full.yaml b/.github/workflows/vllm_ascend_test_full.yaml
index 6306032..5f5089d 100644
--- a/.github/workflows/vllm_ascend_test_full.yaml
+++ b/.github/workflows/vllm_ascend_test_full.yaml
@@ -68,7 +68,7 @@ jobs:
     name: e2e-full
     strategy:
       matrix:
-        vllm_version: [17b4c6685ce62d5652654784d6771a3d38e4273e, v0.10.2]
+        vllm_version: [releases/v0.11.0, v0.10.2]
     needs: [changes]
     if: ${{ needs.changes.outputs.e2e_tracker == 'true' }}
     uses: ./.github/workflows/_e2e_test.yaml
diff --git a/vllm_ascend/models/qwen2_5_vl.py b/vllm_ascend/models/qwen2_5_vl.py
index 4497104..a3e53c0 100644
--- a/vllm_ascend/models/qwen2_5_vl.py
+++ b/vllm_ascend/models/qwen2_5_vl.py
@@ -42,6 +42,8 @@ from vllm.model_executor.models.qwen2_5_vl import (
 from vllm.model_executor.models.utils import maybe_prefix
 from vllm.multimodal import MULTIMODAL_REGISTRY
 
+from vllm_ascend.utils import vllm_version_is
+
 MIN_PAD_SIZE = 64  # min_size to pad weight
 MAX_PAD_SIZE = 128  # max_size to pad weight
 
@@ -496,12 +498,20 @@ class AscendQwen2_5_VLForConditionalGeneration(
         super().__init__(vllm_config=vllm_config, prefix=prefix)
         config: Qwen2_5_VLConfig = vllm_config.model_config.hf_config
         quant_config = vllm_config.quant_config
-        self.visual = AscendQwen2_5_VisionTransformer(
-            vision_config=config.vision_config,
-            norm_eps=getattr(config, "rms_norm_eps", 1e-6),
-            quant_config=self._maybe_ignore_quant_config(quant_config),
-            prefix=maybe_prefix(prefix, "visual"),
-        )
+        if vllm_version_is("0.10.2"):
+            self.visual = AscendQwen2_5_VisionTransformer(
+                vision_config=config.vision_config,
+                norm_eps=getattr(config, "rms_norm_eps", 1e-6),
+                quant_config=self._maybe_ignore_quant_config(quant_config),
+                prefix=maybe_prefix(prefix, "visual"),
+            )
+        else:
+            self.visual = AscendQwen2_5_VisionTransformer(
+                vision_config=config.vision_config,
+                norm_eps=getattr(config, "rms_norm_eps", 1e-6),
+                quant_config=self.quant_config,
+                prefix=maybe_prefix(prefix, "visual"),
+            )
 
     def _process_image_input(self, image_input) -> tuple[torch.Tensor, ...]:
 
diff --git a/vllm_ascend/models/qwen2_5_vl_without_padding.py b/vllm_ascend/models/qwen2_5_vl_without_padding.py
index d3fac4c..54f5947 100644
--- a/vllm_ascend/models/qwen2_5_vl_without_padding.py
+++ b/vllm_ascend/models/qwen2_5_vl_without_padding.py
@@ -68,6 +68,7 @@ from vllm.model_executor.models.utils import WeightsMapper, maybe_prefix
 from vllm.multimodal import MULTIMODAL_REGISTRY
 
 from vllm_ascend.models.qwen2_5_vl import AscendQwen2_5_VisionRotaryEmbedding
+from vllm_ascend.utils import vllm_version_is
 
 
 class AscendQwen2_5_VisionAttention_Without_Padding(Qwen2_5_VisionAttention):
@@ -483,12 +484,20 @@ class AscendQwen2_5_VLForConditionalGeneration_Without_Padding(
         super().__init__(vllm_config=vllm_config, prefix=prefix)
         config: Qwen2_5_VLConfig = vllm_config.model_config.hf_config
         quant_config = vllm_config.quant_config
-        self.visual = AscendQwen2_5_VisionTransformer_Without_Padding(
-            vision_config=config.vision_config,
-            norm_eps=getattr(config, "rms_norm_eps", 1e-6),
-            quant_config=self._maybe_ignore_quant_config(quant_config),
-            prefix=maybe_prefix(prefix, "visual"),
-        )
+        if vllm_version_is("0.10.2"):
+            self.visual = AscendQwen2_5_VisionTransformer_Without_Padding(
+                vision_config=config.vision_config,
+                norm_eps=getattr(config, "rms_norm_eps", 1e-6),
+                quant_config=self._maybe_ignore_quant_config(quant_config),
+                prefix=maybe_prefix(prefix, "visual"),
+            )
+        else:
+            self.visual = AscendQwen2_5_VisionTransformer_Without_Padding(
+                vision_config=config.vision_config,
+                norm_eps=getattr(config, "rms_norm_eps", 1e-6),
+                quant_config=self.quant_config,
+                prefix=maybe_prefix(prefix, "visual"),
+            )
 
     def _process_image_input(self, image_input) -> tuple[torch.Tensor, ...]:
 
@@ -554,12 +563,20 @@ class AscendQwen3VLForConditionalGeneration(Qwen3VLForConditionalGeneration):
         super().__init__(vllm_config=vllm_config, prefix=prefix)
         config: Qwen3VLConfig = vllm_config.model_config.hf_config
         quant_config = vllm_config.quant_config
-        self.visual = AscendQwen3_VisionTransformer(
-            config.vision_config,
-            norm_eps=getattr(config, "rms_norm_eps", 1e-6),
-            quant_config=self._maybe_ignore_quant_config(quant_config),
-            prefix=maybe_prefix(prefix, "visual"),
-            use_data_parallel=self.use_data_parallel)
+        if vllm_version_is("0.10.2"):
+            self.visual = AscendQwen3_VisionTransformer(
+                config.vision_config,
+                norm_eps=getattr(config, "rms_norm_eps", 1e-6),
+                quant_config=self._maybe_ignore_quant_config(quant_config),
+                prefix=maybe_prefix(prefix, "visual"),
+                use_data_parallel=self.use_data_parallel)
+        else:
+            self.visual = AscendQwen3_VisionTransformer(
+                config.vision_config,
+                norm_eps=getattr(config, "rms_norm_eps", 1e-6),
+                quant_config=self.quant_config,
+                prefix=maybe_prefix(prefix, "visual"),
+                use_data_parallel=self.use_data_parallel)
 
 
 @MULTIMODAL_REGISTRY.register_processor(Qwen3VLMultiModalProcessor,
@@ -596,11 +613,19 @@ class AscendQwen3VLMoeForConditionalGeneration(
         multimodal_config = vllm_config.model_config.multimodal_config
         self.multimodal_config = multimodal_config
         self.use_data_parallel = multimodal_config.mm_encoder_tp_mode == "data"
-
-        self.visual = AscendQwen3_VisionTransformer(
-            config.vision_config,
-            norm_eps=getattr(config, "rms_norm_eps", 1e-6),
-            quant_config=self._maybe_ignore_quant_config(quant_config),
-            prefix=maybe_prefix(prefix, "visual"),
-            use_data_parallel=self.use_data_parallel,
-        )
\ No newline at end of file
+        if vllm_version_is("0.10.2"):
+            self.visual = AscendQwen3_VisionTransformer(
+                config.vision_config,
+                norm_eps=getattr(config, "rms_norm_eps", 1e-6),
+                quant_config=self._maybe_ignore_quant_config(quant_config),
+                prefix=maybe_prefix(prefix, "visual"),
+                use_data_parallel=self.use_data_parallel,
+            )
+        else:
+            self.visual = AscendQwen3_VisionTransformer(
+                config.vision_config,
+                norm_eps=getattr(config, "rms_norm_eps", 1e-6),
+                quant_config=self.quant_config,
+                prefix=maybe_prefix(prefix, "visual"),
+                use_data_parallel=self.use_data_parallel,
+            )
diff --git a/vllm_ascend/models/qwen2_vl.py b/vllm_ascend/models/qwen2_vl.py
index a677b06..0eea6f8 100644
--- a/vllm_ascend/models/qwen2_vl.py
+++ b/vllm_ascend/models/qwen2_vl.py
@@ -40,6 +40,8 @@ from vllm.model_executor.models.qwen2_vl import (
 from vllm.model_executor.models.utils import maybe_prefix
 from vllm.multimodal import MULTIMODAL_REGISTRY
 
+from vllm_ascend.utils import vllm_version_is
+
 MIN_PAD_SIZE = 64  # min_size to pad weight
 MAX_PAD_SIZE = 128  # max_size to pad weight
 
@@ -343,10 +345,18 @@ class AscendQwen2VLForConditionalGeneration(Qwen2VLForConditionalGeneration):
 
     def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
         super().__init__(vllm_config=vllm_config, prefix=prefix)
-        self.visual = AscendQwen2VisionTransformer(
-            self.config.vision_config,
-            norm_eps=getattr(self.config, "rms_norm_eps", 1e-6),
-            quant_config=self._maybe_ignore_quant_config(
-                vllm_config.quant_config),
-            prefix=maybe_prefix(prefix, "visual"),
-        )
\ No newline at end of file
+        if vllm_version_is("0.10.2"):
+            self.visual = AscendQwen2VisionTransformer(
+                self.config.vision_config,
+                norm_eps=getattr(self.config, "rms_norm_eps", 1e-6),
+                quant_config=self._maybe_ignore_quant_config(
+                    vllm_config.quant_config),
+                prefix=maybe_prefix(prefix, "visual"),
+            )
+        else:
+            self.visual = AscendQwen2VisionTransformer(
+                self.config.vision_config,
+                norm_eps=getattr(self.config, "rms_norm_eps", 1e-6),
+                quant_config=self.vllm_config.quant_config,
+                prefix=maybe_prefix(prefix, "visual"),
+            )