diff --git a/setup.py b/setup.py
index b0c54cdf..a38d9563 100644
--- a/setup.py
+++ b/setup.py
@@ -135,23 +135,22 @@ else:
 def gen_build_info():
     soc_version = envs.SOC_VERSION
 
-    # TODO(zzzzwwjj): Add A5 case
     soc_to_device = {
-        "910b": "_910B",
-        "910c": "_910_93",
+        "910b": "A2",
+        "910c": "A3",
         "310p": "_310P",
-        "ascend910b1": "_910B",
-        "ascend910b2": "_910B",
-        "ascend910b2c": "_910B",
-        "ascend910b3": "_910B",
-        "ascend910b4": "_910B",
-        "ascend910b4-1": "_910B",
-        "ascend910_9391": "_910_93",
-        "ascend910_9381": "_910_93",
-        "ascend910_9372": "_910_93",
-        "ascend910_9392": "_910_93",
-        "ascend910_9382": "_910_93",
-        "ascend910_9362": "_910_93",
+        "ascend910b1": "A2",
+        "ascend910b2": "A2",
+        "ascend910b2c": "A2",
+        "ascend910b3": "A2",
+        "ascend910b4": "A2",
+        "ascend910b4-1": "A2",
+        "ascend910_9391": "A3",
+        "ascend910_9381": "A3",
+        "ascend910_9372": "A3",
+        "ascend910_9392": "A3",
+        "ascend910_9382": "A3",
+        "ascend910_9362": "A3",
         "ascend310p1": "_310P",
         "ascend310p3": "_310P",
         "ascend310p5": "_310P",
@@ -160,7 +159,7 @@ def gen_build_info():
         "ascend310p3vir02": "_310P",
         "ascend310p3vir04": "_310P",
         "ascend310p3vir08": "_310P",
-        "ascend910_9579": "_910_95",
+        "ascend910_9579": "A5",
     }
 
     assert soc_version in soc_to_device, f"Undefined soc_version: {soc_version}. Please file an issue to vllm-ascend."
diff --git a/tests/e2e/multicard/test_aclgraph_capture_replay.py b/tests/e2e/multicard/test_aclgraph_capture_replay.py
index e81b5615..a5e3b846 100644
--- a/tests/e2e/multicard/test_aclgraph_capture_replay.py
+++ b/tests/e2e/multicard/test_aclgraph_capture_replay.py
@@ -215,7 +215,7 @@ def test_aclgraph_capture_replay_metrics_dp2(
     # Part A: Warmup runs (Profile run + 2 runs per captured graph)
     warmup_runs = 1 + (2 * max_batch_sizes)
     soc_version = get_ascend_device_type()
-    if soc_version in {AscendDeviceType._910_93} and "DeepSeek" in model:
+    if soc_version in {AscendDeviceType.A3} and "DeepSeek" in model:
         # An extra warmup run is needed for MC2 warmup here
         warmup_runs += 1
 
diff --git a/tests/ut/attention/test_attention_v1.py b/tests/ut/attention/test_attention_v1.py
index 5923a021..e2c83e41 100644
--- a/tests/ut/attention/test_attention_v1.py
+++ b/tests/ut/attention/test_attention_v1.py
@@ -26,7 +26,7 @@ class TestAscendAttentionBackend(TestBase):
                          AscendAttentionMetadataBuilder)
 
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     def test_get_kv_cache_shape_not_310p(self, mock_soc_version):
         result = AscendAttentionBackend.get_kv_cache_shape(10, 20, 30, 40)
         self.assertEqual(result, (2, 10, 20, 30, 40))
@@ -103,7 +103,7 @@ class TestAscendAttentionMetadataBuilder(TestBase):
 
     @patch('vllm_ascend.attention.attention_v1.AscendMetadata')
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     def test_build_non_310p(self, mock_soc_version, mock_ascend_metadata):
         common_attn_metadata = AscendCommonAttentionMetadata(
             query_start_loc=torch.tensor([0, 2, 5, 9]),
diff --git a/tests/ut/ops/test_activation.py b/tests/ut/ops/test_activation.py
index 12e8f4c2..9b802365 100644
--- a/tests/ut/ops/test_activation.py
+++ b/tests/ut/ops/test_activation.py
@@ -49,7 +49,7 @@ def test_SiluAndMul_forward(mock_maybe_prefetch_mlp_down_proj,
 
     with patch("vllm_ascend.utils.get_ascend_device_type",
                return_value=AscendDeviceType._310P
-               if is_310p else AscendDeviceType._910_93):
+               if is_310p else AscendDeviceType.A3):
         layer = SiluAndMul()
         out = layer.forward(dummy_tensor)
 
diff --git a/tests/ut/ops/test_fused_moe.py b/tests/ut/ops/test_fused_moe.py
index d1981b2a..94a52b7c 100644
--- a/tests/ut/ops/test_fused_moe.py
+++ b/tests/ut/ops/test_fused_moe.py
@@ -127,7 +127,7 @@ def mock_dist_env(mocker: MockerFixture):
             return_value=mock_forward_context_obj), \
         patch('vllm_ascend.ops.fused_moe.prepare_finalize.get_forward_context',
             return_value=mock_forward_context_obj), \
-        patch("vllm_ascend.utils.get_ascend_device_type", return_value=AscendDeviceType._910_93), \
+        patch("vllm_ascend.utils.get_ascend_device_type", return_value=AscendDeviceType.A3), \
         patch('vllm_ascend.ops.fused_moe.moe_mlp.get_forward_context',
                 return_value=mock_forward_context_obj), \
         patch('vllm_ascend.ops.fused_moe.moe_comm_method.MC2CommImpl._get_token_dispatcher',
@@ -323,7 +323,7 @@ class TestUnifiedApplyMLP(TestBase):
 
     @patch('vllm_ascend.ops.fused_moe.moe_mlp.get_forward_context')
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     @patch('torch_npu.npu_grouped_matmul')
     @patch('torch_npu.npu_dynamic_quant')
     @patch('torch_npu.npu_dequant_swiglu_quant')
@@ -386,7 +386,7 @@ class TestUnifiedApplyMLP(TestBase):
         self.assertEqual(result.dtype, torch.bfloat16)
 
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     @patch('torch_npu.npu_grouped_matmul')
     @patch('torch_npu.npu_swiglu')
     @patch('torch_npu.npu_dynamic_quant')
diff --git a/tests/ut/ops/test_layernorm.py b/tests/ut/ops/test_layernorm.py
index e50656e8..03befc7e 100644
--- a/tests/ut/ops/test_layernorm.py
+++ b/tests/ut/ops/test_layernorm.py
@@ -30,7 +30,7 @@ def test_RMSNorm_forward(mock_add_rmsnorm, mock_rmsnorm, is_310p, residual,
 
     with patch("vllm_ascend.utils.get_ascend_device_type",
                return_value=AscendDeviceType._310P
-               if is_310p else AscendDeviceType._910_93):
+               if is_310p else AscendDeviceType.A3):
         layer = RMSNorm(hidden_size=8, eps=1e-05)
         if residual is not None:
             out_x, out_residual = layer.forward_oot(dummy_tensor, residual)
diff --git a/tests/ut/ops/test_rotary_embedding.py b/tests/ut/ops/test_rotary_embedding.py
index f3e263ff..569b70ab 100644
--- a/tests/ut/ops/test_rotary_embedding.py
+++ b/tests/ut/ops/test_rotary_embedding.py
@@ -99,7 +99,7 @@ class TestAscendRotaryEmbedding(unittest.TestCase):
 
     @patch('torch.ops._C_ascend')
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     @patch('vllm_ascend.ops.rotary_embedding._custom_rotary_embedding_enabled',
            return_value=True)
     @patch('torch.ops._npu_rotary_embedding')
diff --git a/tests/ut/ops/test_token_dispatcher.py b/tests/ut/ops/test_token_dispatcher.py
index 26a741c3..140bae5c 100644
--- a/tests/ut/ops/test_token_dispatcher.py
+++ b/tests/ut/ops/test_token_dispatcher.py
@@ -53,7 +53,7 @@ class TestTokenDispatcherWithMC2(TestBase):
         # Mock get_ascend_device_type()
         self.ascend_soc_version_patch = patch(
             "vllm_ascend.ops.fused_moe.token_dispatcher.get_ascend_device_type",
-            return_value=AscendDeviceType._910_93)
+            return_value=AscendDeviceType.A3)
         self.ascend_soc_version_patch.start()
 
         kwargs = {"with_quant": False, "top_k": 8, "num_experts": 128}
diff --git a/tests/ut/quantization/test_w8a8.py b/tests/ut/quantization/test_w8a8.py
index c574d998..9a1c0e8d 100644
--- a/tests/ut/quantization/test_w8a8.py
+++ b/tests/ut/quantization/test_w8a8.py
@@ -347,7 +347,7 @@ class TestAscendC8KVCacheMethod(TestBase):
             self.assertEqual(param.shape, expected_shape)
 
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     def test_process_weights_after_loading_not_310p(self, mock_soc_version):
         key_data = torch.ones(4 * 64)
         value_data = torch.ones(4 * 64) * 2
diff --git a/tests/ut/test_platform.py b/tests/ut/test_platform.py
index 6a2ce5c8..a5257ce9 100644
--- a/tests/ut/test_platform.py
+++ b/tests/ut/test_platform.py
@@ -231,7 +231,7 @@ class TestNPUPlatform(TestBase):
     @patch("vllm_ascend.ascend_config.init_ascend_config")
     @patch("vllm_ascend.utils.update_aclgraph_sizes")
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     @patch("os.environ", {})
     @patch(
         "vllm_ascend.core.recompute_scheduler.RecomputeSchedulerConfig.initialize_from_config"
@@ -263,7 +263,7 @@ class TestNPUPlatform(TestBase):
         mock_init_ascend.assert_called_once_with(vllm_config)
 
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     @patch("vllm_ascend.ascend_config.init_ascend_config")
     @patch(
         "vllm_ascend.core.recompute_scheduler.RecomputeSchedulerConfig.initialize_from_config"
@@ -288,7 +288,7 @@ class TestNPUPlatform(TestBase):
         self.assertTrue("Model config is missing" in cm.output[0])
 
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     @patch("vllm_ascend.ascend_config.init_ascend_config")
     @patch(
         "vllm_ascend.core.recompute_scheduler.RecomputeSchedulerConfig.initialize_from_config"
@@ -324,7 +324,7 @@ class TestNPUPlatform(TestBase):
         )
 
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     @patch("vllm_ascend.utils.update_default_aclgraph_sizes")
     @patch("vllm_ascend.ascend_config.init_ascend_config")
     @patch(
@@ -365,7 +365,7 @@ class TestNPUPlatform(TestBase):
     @pytest.mark.skip(
         "Revert me when vllm support setting cudagraph_mode on oot platform")
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     @patch("vllm_ascend.ascend_config.init_ascend_config")
     def test_check_and_update_config_unsupported_cudagraph_mode(
             self, mock_init_ascend, mock_soc_version):
@@ -394,7 +394,7 @@ class TestNPUPlatform(TestBase):
             )
 
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     @patch("vllm_ascend.ascend_config.init_ascend_config")
     @patch(
         "vllm_ascend.core.recompute_scheduler.RecomputeSchedulerConfig.initialize_from_config"
@@ -421,7 +421,7 @@ class TestNPUPlatform(TestBase):
         self.assertEqual(vllm_config.cache_config.block_size, 128)
 
     @patch('vllm_ascend.utils.get_ascend_device_type',
-           return_value=AscendDeviceType._910_93)
+           return_value=AscendDeviceType.A3)
     @patch("vllm_ascend.ascend_config.init_ascend_config")
     @patch(
         "vllm_ascend.core.recompute_scheduler.RecomputeSchedulerConfig.initialize_from_config"
diff --git a/vllm_ascend/ascend_forward_context.py b/vllm_ascend/ascend_forward_context.py
index 3ebf11b4..ebc22bd1 100644
--- a/vllm_ascend/ascend_forward_context.py
+++ b/vllm_ascend/ascend_forward_context.py
@@ -244,7 +244,7 @@ def select_moe_comm_method(num_tokens: int,
 
     if not vllm_config.parallel_config.enable_expert_parallel:
         moe_comm_type = MoECommType.ALLGATHER
-    elif soc_version in {AscendDeviceType._910B}:
+    elif soc_version in {AscendDeviceType.A2}:
         if (num_tokens <= mc2_tokens_capacity
                 and vllm_config.parallel_config.world_size_across_dp /
                 vllm_config.parallel_config.pipeline_parallel_size >= 16):
@@ -256,7 +256,7 @@ def select_moe_comm_method(num_tokens: int,
             else:
                 moe_comm_type = MoECommType.ALLGATHER
 
-    elif soc_version in {AscendDeviceType._910_93}:
+    elif soc_version in {AscendDeviceType.A3}:
         ascend_config = get_ascend_config()
         dynamic_eplb = ascend_config.dynamic_eplb or ascend_config.expert_map_record_path
         # TODO: drop the EP-size guard when dispatch_ffn_combine supports larger EP sizes
diff --git a/vllm_ascend/attention/attention_v1.py b/vllm_ascend/attention/attention_v1.py
index 7faa30a9..875dd432 100644
--- a/vllm_ascend/attention/attention_v1.py
+++ b/vllm_ascend/attention/attention_v1.py
@@ -640,7 +640,7 @@ class AscendAttentionBackendImpl(AttentionImpl):
         attn_metadata: AscendMetadata,
         output: Optional[torch.Tensor] = None,
     ) -> torch.Tensor:
-        if get_ascend_device_type() == AscendDeviceType._910_95:
+        if get_ascend_device_type() == AscendDeviceType.A5:
             return self._forward_decode_only_ascend91095(
                 query, attn_metadata, output)
         if self.sliding_window is not None and attn_metadata.seq_lens.shape[
@@ -729,7 +729,7 @@ class AscendAttentionBackendImpl(AttentionImpl):
             if self.key_cache is None:
                 self.key_cache, self.value_cache = kv_cache[0], kv_cache[1]
             slots = attn_metadata.slot_mapping
-            if get_ascend_device_type() == AscendDeviceType._910_95:
+            if get_ascend_device_type() == AscendDeviceType.A5:
                 # TODO: Once eagle running to here, it may has error because of the 0 dim of slot_mapping.
                 # Should check if the 0 dim of slot_mapping must equal to the 0 dim of key.
                 # If it's necessary, the slots should be sliced.
diff --git a/vllm_ascend/ops/fused_moe/token_dispatcher.py b/vllm_ascend/ops/fused_moe/token_dispatcher.py
index e45504d9..1246d648 100644
--- a/vllm_ascend/ops/fused_moe/token_dispatcher.py
+++ b/vllm_ascend/ops/fused_moe/token_dispatcher.py
@@ -99,7 +99,7 @@ class TokenDispatcherWithMC2(MoETokenDispatcher):
         self.enable_dispatch_v2 = hasattr(torch_npu,
                                           "npu_moe_distribute_dispatch_v2")
         self.need_extra_args = (
-            get_ascend_device_type() == AscendDeviceType._910_93)
+            get_ascend_device_type() == AscendDeviceType.A3)
 
         # NOTE: When in A2, setting the environment variables HCCL_INTRA_PCIE_ENABLE=1 and
         # HCCL_INTRA_ROCE_ENABLE=0 can reduce cross-machine communication traffic and significantly
diff --git a/vllm_ascend/ops/rotary_embedding.py b/vllm_ascend/ops/rotary_embedding.py
index 7f860470..566c48f4 100644
--- a/vllm_ascend/ops/rotary_embedding.py
+++ b/vllm_ascend/ops/rotary_embedding.py
@@ -499,7 +499,7 @@ class AscendMRotaryEmbedding(MRotaryEmbedding):
         key: torch.Tensor,
     ):
         if self.mrope_section != [16, 24, 24] or \
-            get_ascend_device_type() == AscendDeviceType._910_95:
+            get_ascend_device_type() == AscendDeviceType.A5:
             return super().forward_oot(positions, query, key)
 
         import torch_npu
diff --git a/vllm_ascend/utils.py b/vllm_ascend/utils.py
index dc02e1bc..9e2431bb 100644
--- a/vllm_ascend/utils.py
+++ b/vllm_ascend/utils.py
@@ -674,10 +674,10 @@ def register_ascend_customop(vllm_config: Optional[VllmConfig] = None):
 
 
 class AscendDeviceType(Enum):
-    _910B = 0  # A2
-    _910_93 = 1  # A3
+    A2 = 0
+    A3 = 1
     _310P = 2
-    _910_95 = 3  # A5
+    A5 = 3
 
 
 _ascend_device_type = None
@@ -696,13 +696,13 @@ def check_ascend_device_type():
 
     soc_version = torch_npu.npu.get_soc_version()
     if 220 <= soc_version <= 225:
-        cur_device_type = AscendDeviceType._910B
+        cur_device_type = AscendDeviceType.A2
     elif 250 <= soc_version <= 255:
-        cur_device_type = AscendDeviceType._910_93
+        cur_device_type = AscendDeviceType.A3
     elif 200 <= soc_version <= 205:
         cur_device_type = AscendDeviceType._310P
     elif soc_version == 260:
-        cur_device_type = AscendDeviceType._910_95
+        cur_device_type = AscendDeviceType.A5
     else:
         raise RuntimeError(f"Can not support soc_version: {soc_version}.")
 
diff --git a/vllm_ascend/worker/worker_v1.py b/vllm_ascend/worker/worker_v1.py
index 29e2fb85..3e1f3f59 100644
--- a/vllm_ascend/worker/worker_v1.py
+++ b/vllm_ascend/worker/worker_v1.py
@@ -90,7 +90,7 @@ class NPUWorker(WorkerBase):
         # Register ops when worker init.
         from vllm_ascend import ops
         ops.register_dummy_fusion_op()
-        if get_ascend_device_type() != AscendDeviceType._910_95:
+        if get_ascend_device_type() != AscendDeviceType.A5:
             _register_atb_extensions()
         register_ascend_customop(vllm_config)
         # init ascend config and soc version
@@ -360,7 +360,7 @@ class NPUWorker(WorkerBase):
             self.model_runner.capture_model()
         # Call ATB matmul to warm up; otherwise, the first operation (ReshapeAndCache)
         # may cause performance degradation at runtime.
-        if get_ascend_device_type() != AscendDeviceType._910_95:
+        if get_ascend_device_type() != AscendDeviceType.A5:
             self._warm_up_atb()
         # Reset the seed to ensure that the random state is not affected by
         # the model initialization and profiling.