diff --git a/vllm/model_executor/models/bailing_moe.py b/vllm/model_executor/models/bailing_moe.py
index 325ba7bbad83..ccfc3997e45c 100644
--- a/vllm/model_executor/models/bailing_moe.py
+++ b/vllm/model_executor/models/bailing_moe.py
@@ -47,7 +47,6 @@
 from vllm.model_executor.layers.quantization.base_config import (
     QuantizationConfig)
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -485,7 +484,6 @@ def __init__(
         else:
             self.lm_head = PPMissingLayer()
 
-        self.sampler = get_sampler()
         self.make_empty_intermediate_tensors = (
             self.model.make_empty_intermediate_tensors)
 
@@ -512,14 +510,6 @@ def compute_logits(
                                        sampling_metadata)
         return logits
 
-    def sample(
-        self,
-        logits: torch.Tensor,
-        sampling_metadata: SamplingMetadata,
-    ) -> Optional[SamplerOutput]:
-        next_tokens = self.sampler(logits, sampling_metadata)
-        return next_tokens
-
     def load_weights(self, weights: Iterable[tuple[str,
                                                    torch.Tensor]]) -> set[str]:
         loader = AutoWeightsLoader(
diff --git a/vllm/model_executor/models/granite_speech.py b/vllm/model_executor/models/granite_speech.py
index 6c7c9f5cc938..6a4dee9ae48d 100644
--- a/vllm/model_executor/models/granite_speech.py
+++ b/vllm/model_executor/models/granite_speech.py
@@ -36,7 +36,6 @@
 from vllm.model_executor.layers.linear import (ColumnParallelLinear,
                                                RowParallelLinear)
 from vllm.model_executor.layers.quantization import QuantizationConfig
-from vllm.model_executor.layers.sampler import get_sampler
 from vllm.model_executor.models.module_mapping import MultiModelKeys
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.multimodal import MULTIMODAL_REGISTRY
@@ -549,7 +548,6 @@ def __init__(self, *, vllm_config: VllmConfig, prefix: str):
         self.config = config
         self.quant_config = quant_config
         self.cache_config = cache_config
-        self.sampler = get_sampler()
 
         # The language model is typically a Granite LLM
         self.language_model = init_vllm_registered_model(
diff --git a/vllm/model_executor/models/hunyuan_v1_moe.py b/vllm/model_executor/models/hunyuan_v1_moe.py
index 89ca3e8a6071..43ffba00721e 100644
--- a/vllm/model_executor/models/hunyuan_v1_moe.py
+++ b/vllm/model_executor/models/hunyuan_v1_moe.py
@@ -49,7 +49,6 @@
 from vllm.model_executor.layers.quantization.base_config import (
     QuantizationConfig)
 from vllm.model_executor.layers.rotary_embedding import get_rope
-from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     DEFAULT_VOCAB_PADDING_SIZE, ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import (
@@ -661,7 +660,6 @@ def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
             self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,
                                                     config.vocab_size,
                                                     logit_scale)
-            self.sampler = get_sampler()
         else:
             self.lm_head = PPMissingLayer()
 
@@ -685,14 +683,6 @@ def compute_logits(
                                        sampling_metadata)
         return logits
 
-    def sample(
-        self,
-        logits: torch.Tensor,
-        sampling_metadata: SamplingMetadata,
-    ) -> Optional[SamplerOutput]:
-        next_tokens = self.sampler(logits, sampling_metadata)
-        return next_tokens
-
     def make_empty_intermediate_tensors(
             self, batch_size: int, dtype: torch.dtype,
             device: torch.device) -> IntermediateTensors:
diff --git a/vllm/model_executor/models/mimo.py b/vllm/model_executor/models/mimo.py
index 9b83f848ef42..5b497dd9d89f 100644
--- a/vllm/model_executor/models/mimo.py
+++ b/vllm/model_executor/models/mimo.py
@@ -36,7 +36,6 @@
 from vllm.distributed import get_pp_group
 from vllm.logger import init_logger
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
-from vllm.model_executor.layers.sampler import get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
 from vllm.model_executor.model_loader.weight_utils import (
     default_weight_loader, maybe_remap_kv_scale_name)
@@ -176,7 +175,6 @@ def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
             self.lm_head = PPMissingLayer()
 
         self.logits_processor = LogitsProcessor(config.vocab_size)
-        self.sampler = get_sampler()
 
         self.make_empty_intermediate_tensors = (
             self.model.make_empty_intermediate_tensors)
diff --git a/vllm/model_executor/models/mimo_mtp.py b/vllm/model_executor/models/mimo_mtp.py
index 6066ec76c5fc..19afc5be3fb8 100644
--- a/vllm/model_executor/models/mimo_mtp.py
+++ b/vllm/model_executor/models/mimo_mtp.py
@@ -30,7 +30,6 @@
 from vllm.model_executor.layers.layernorm import RMSNorm
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
-from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
@@ -161,8 +160,6 @@ def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
         self.lm_head = ParallelLMHead(self.config.vocab_size,
                                       self.config.hidden_size)
 
-        self.sampler = get_sampler()
-
     def forward(
         self,
         input_ids: torch.Tensor,
@@ -187,14 +184,6 @@ def compute_logits(
         return self.model.compute_logits(hidden_states, self.lm_head,
                                          sampling_metadata, spec_step_idx)
 
-    def sample(
-        self,
-        logits: torch.Tensor,
-        sampling_metadata: SamplingMetadata,
-    ) -> Optional[SamplerOutput]:
-        next_tokens = self.sampler(logits, sampling_metadata)
-        return next_tokens
-
     def load_weights(self, weights: Iterable[tuple[str,
                                                    torch.Tensor]]) -> set[str]:
         stacked_params_mapping = [
diff --git a/vllm/model_executor/models/phi4flash.py b/vllm/model_executor/models/phi4flash.py
index c1dd9fab7faa..a4ded2b7a304 100644
--- a/vllm/model_executor/models/phi4flash.py
+++ b/vllm/model_executor/models/phi4flash.py
@@ -23,7 +23,6 @@
     causal_conv1d_fn, causal_conv1d_update)
 from vllm.model_executor.layers.mamba.ops.mamba_ssm import (
     selective_scan_fn, selective_state_update)
-from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     DEFAULT_VOCAB_PADDING_SIZE, ParallelLMHead, VocabParallelEmbedding)
 from vllm.model_executor.models.interfaces import (HasInnerState, IsHybrid,
@@ -641,7 +640,6 @@ def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
         self.logits_processor = LogitsProcessor(self.unpadded_vocab_size,
                                                 config.vocab_size,
                                                 logits_as_input=False)
-        self.sampler = get_sampler()
 
     def forward(
         self,
@@ -709,14 +707,6 @@ def compute_logits(
             prune_hidden_states=prune_hidden_states)
         return processed_logits
 
-    def sample(
-        self,
-        logits: torch.Tensor,
-        sampling_metadata: SamplingMetadata,
-    ) -> Optional[SamplerOutput]:
-        next_tokens = self.sampler(logits, sampling_metadata)
-        return next_tokens
-
     def load_weights(
         self,
         weights: Iterable[tuple[str, torch.Tensor]],