diff --git a/tests/plugins/bge_m3_sparse_plugin/bge_m3_sparse_processor/sparse_embeddings_processor.py b/tests/plugins/bge_m3_sparse_plugin/bge_m3_sparse_processor/sparse_embeddings_processor.py
index 74cf122503..b67d1668ff 100644
--- a/tests/plugins/bge_m3_sparse_plugin/bge_m3_sparse_processor/sparse_embeddings_processor.py
+++ b/tests/plugins/bge_m3_sparse_plugin/bge_m3_sparse_processor/sparse_embeddings_processor.py
@@ -58,13 +58,11 @@ class BgeM3SparseEmbeddingsProcessor(
         params.dimensions = self.embed_dimensions
         return params
 
-    def parse_request(
-        self, request_data: object
-    ) -> SparseEmbeddingCompletionRequestMixin:
+    def parse_data(self, data: object) -> SparseEmbeddingCompletionRequestMixin:
         # for vllm.entrypoints.llm.LLM, offline mode, calls `encode` directly.
-        if isinstance(request_data, dict):
-            return SparseEmbeddingCompletionRequestMixin(**request_data)
-        raise TypeError("request_data should be a dictionary")
+        if isinstance(data, dict):
+            return SparseEmbeddingCompletionRequestMixin(**data)
+        raise TypeError("data should be a dictionary")
 
     def pre_process(
         self,
diff --git a/vllm/config/attention.py b/vllm/config/attention.py
index ea92be32cd..d5a150c92a 100644
--- a/vllm/config/attention.py
+++ b/vllm/config/attention.py
@@ -119,7 +119,7 @@ class AttentionConfig:
         if self.use_fp4_indexer_cache is not None:
             logger.warning(
                 "use_fp4_indexer_cache is deprecated and will be removed in "
-                "v0.19. Use indexer_kv_dtype instead (True -> 'mxfp4')."
+                "v0.29. Use indexer_kv_dtype instead (True -> 'mxfp4')."
             )
             if self.use_fp4_indexer_cache:
                 if self.indexer_kv_dtype not in ("auto", "mxfp4"):
diff --git a/vllm/entrypoints/pooling/base/io_processor.py b/vllm/entrypoints/pooling/base/io_processor.py
index 940987fc16..05ec69e44a 100644
--- a/vllm/entrypoints/pooling/base/io_processor.py
+++ b/vllm/entrypoints/pooling/base/io_processor.py
@@ -213,7 +213,7 @@ class PoolingIOProcessor:
             if param.task is None:
                 param.task = pooling_task
             elif pooling_task == "plugin":
-                # `plugin` task uses io_processor.parse_request to verify inputs.
+                # `plugin` task uses io_processor.parse_data to verify inputs.
                 # We actually allow plugin to overwrite pooling_task.
                 pass
             elif param.task != pooling_task:
diff --git a/vllm/entrypoints/pooling/offline.py b/vllm/entrypoints/pooling/offline.py
index ade3c03984..f9def808de 100644
--- a/vllm/entrypoints/pooling/offline.py
+++ b/vllm/entrypoints/pooling/offline.py
@@ -173,7 +173,7 @@ class PoolingOfflineMixin(OfflineInferenceMixin):
                 "Try converting the model using `--convert classify`."
             )
 
-        # plugin task uses io_processor.parse_request to verify inputs
+        # plugin task uses io_processor.parse_data to verify inputs
         if pooling_task != "plugin" and pooling_task != self.pooling_task:
             if pooling_task not in self.supported_tasks:
                 raise ValueError(
diff --git a/vllm/entrypoints/pooling/pooling/serving.py b/vllm/entrypoints/pooling/pooling/serving.py
index 218052d19e..56bc7c7865 100644
--- a/vllm/entrypoints/pooling/pooling/serving.py
+++ b/vllm/entrypoints/pooling/pooling/serving.py
@@ -74,7 +74,7 @@ class ServingPooling(PoolingBaseServing):
         assert request.task is not None
         pooling_task = request.task
 
-        # plugin task uses io_processor.parse_request to verify inputs
+        # plugin task uses io_processor.parse_data to verify inputs
         if pooling_task != "plugin" and pooling_task != self.pooling_task:
             if pooling_task not in self.supported_tasks:
                 raise VLLMValidationError(
diff --git a/vllm/envs.py b/vllm/envs.py
index 16c57e7d7e..42643b32ee 100755
--- a/vllm/envs.py
+++ b/vllm/envs.py
@@ -91,8 +91,6 @@ if TYPE_CHECKING:
     VLLM_FLOAT32_MATMUL_PRECISION: Literal["highest", "high", "medium"] = "highest"
     VLLM_BATCH_INVARIANT: bool = False
     VLLM_TRITON_USE_TD: bool | None = None
-    # Deprecated alias of VLLM_TRITON_USE_TD (removed in v0.25).
-    VLLM_TRITON_ATTN_USE_TD: bool | None = None
     VLLM_GPU_SYNC_CHECK: Literal["warn", "error"] | None = None
     MAX_JOBS: str | None = None
     NVCC_THREADS: str | None = None
@@ -563,19 +561,6 @@ def get_env_or_set_default(
 logger = logging.getLogger(__name__)
 
 
-def _deprecated_triton_attn_use_td() -> None:
-    """Warn that VLLM_TRITON_ATTN_USE_TD was renamed to VLLM_TRITON_USE_TD.
-
-    The old name is ignored; VLLM_TRITON_USE_TD is the supported variable.
-    """
-    if "VLLM_TRITON_ATTN_USE_TD" in os.environ:
-        logger.warning(
-            "VLLM_TRITON_ATTN_USE_TD is deprecated and will be removed in "
-            "v0.25. Use VLLM_TRITON_USE_TD instead."
-        )
-    return None
-
-
 def _resolve_rust_cli_path() -> str | None:
     """Resolve the vllm-rs binary path.
 
@@ -651,10 +636,6 @@ environment_variables: dict[str, Callable[[], Any]] = {
     "VLLM_GPU_SYNC_CHECK": env_with_choices(
         "VLLM_GPU_SYNC_CHECK", None, ["warn", "error"]
     ),
-    # Deprecated: renamed to VLLM_TRITON_USE_TD.  Kept registered so it does
-    # not trip the unknown-env-var check; warns on use and is otherwise
-    # ignored.
-    "VLLM_TRITON_ATTN_USE_TD": lambda: _deprecated_triton_attn_use_td(),
     # Maximum number of compilation jobs to run in parallel.
     # By default this is the number of CPUs
     "MAX_JOBS": lambda: os.getenv("MAX_JOBS", None),
diff --git a/vllm/model_executor/models/mistral_large_3.py b/vllm/model_executor/models/mistral_large_3.py
index 603ce5c0f0..4880cda7af 100644
--- a/vllm/model_executor/models/mistral_large_3.py
+++ b/vllm/model_executor/models/mistral_large_3.py
@@ -15,7 +15,7 @@ class MistralLarge3ForCausalLM(DeepseekV3ForCausalLM):
     # (\A...\Z) and after substitution the resulting key always starts with
     # "model." or "lm_head.", so no later pattern can accidentally match again.
     hf_to_vllm_mapper = WeightsMapper(
-        orig_to_new_regex={  # noqa: B950
+        orig_to_new_regex={
             regex.compile(
                 r"\Alayers\.(\d+)\.attention_norm\.weight\Z"
             ): r"model.layers.\1.input_layernorm.weight",
diff --git a/vllm/multimodal/utils.py b/vllm/multimodal/utils.py
index ac9b028b46..85f8f81262 100644
--- a/vllm/multimodal/utils.py
+++ b/vllm/multimodal/utils.py
@@ -12,7 +12,6 @@ from typing import TYPE_CHECKING, Any
 import numpy as np
 import numpy.typing as npt
 from PIL import Image
-from typing_extensions import deprecated
 
 from vllm.inputs import MultiModalPlaceholders
 from vllm.utils.import_utils import LazyLoader
@@ -325,19 +324,6 @@ def group_and_batch_mm_kwargs(
             yield modality, num_items, mm_kwargs_batch
 
 
-@deprecated(
-    "`group_mm_kwargs_by_modality` has been renamed to `group_and_batch_mm_kwargs`. "
-    "The old name will be removed in v0.19."
-)
-def group_mm_kwargs_by_modality(
-    mm_kwargs: list[tuple[str, MultiModalKwargsItem]],
-    *,
-    device: torch.types.Device = None,
-    pin_memory: bool = False,
-) -> Generator[tuple[str, int, BatchedTensorInputs], None, None]:
-    return group_and_batch_mm_kwargs(mm_kwargs, device=device, pin_memory=pin_memory)
-
-
 def fetch_audio(
     audio_url: str,
     audio_io_kwargs: dict[str, Any] | None = None,
diff --git a/vllm/plugins/io_processors/interface.py b/vllm/plugins/io_processors/interface.py
index d75ac8e941..15ba2ba881 100644
--- a/vllm/plugins/io_processors/interface.py
+++ b/vllm/plugins/io_processors/interface.py
@@ -1,6 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-import warnings
 from abc import ABC, abstractmethod
 from collections.abc import AsyncGenerator, Sequence
 from typing import Generic, TypeVar
@@ -25,61 +24,18 @@ class IOProcessor(ABC, Generic[IOProcessorInput, IOProcessorOutput]):
         self.vllm_config = vllm_config
 
     def parse_data(self, data: object) -> IOProcessorInput:
-        if callable(parse_request := getattr(self, "parse_request", None)):
-            warnings.warn(
-                "`parse_request` has been renamed to `parse_data`. "
-                "Please update your IO Processor Plugin to use the new name. "
-                "The old name will be removed in v0.19.",
-                DeprecationWarning,
-                stacklevel=2,
-            )
-
-            return parse_request(data)  # type: ignore
-
         raise NotImplementedError
 
     def merge_sampling_params(
         self,
         params: SamplingParams | None = None,
     ) -> SamplingParams:
-        if callable(
-            validate_or_generate_params := getattr(
-                self, "validate_or_generate_params", None
-            )
-        ):
-            warnings.warn(
-                "`validate_or_generate_params` has been split into "
-                "`merge_sampling_params` and `merge_pooling_params`."
-                "Please update your IO Processor Plugin to use the new methods. "
-                "The old name will be removed in v0.19.",
-                DeprecationWarning,
-                stacklevel=2,
-            )
-
-            return validate_or_generate_params(params)  # type: ignore
-
         return params or SamplingParams()
 
     def merge_pooling_params(
         self,
         params: PoolingParams | None = None,
     ) -> PoolingParams:
-        if callable(
-            validate_or_generate_params := getattr(
-                self, "validate_or_generate_params", None
-            )
-        ):
-            warnings.warn(
-                "`validate_or_generate_params` has been split into "
-                "`merge_sampling_params` and `merge_pooling_params`."
-                "Please update your IO Processor Plugin to use the new methods. "
-                "The old name will be removed in v0.19.",
-                DeprecationWarning,
-                stacklevel=2,
-            )
-
-            return validate_or_generate_params(params)  # type: ignore
-
         return params or PoolingParams(task="plugin")
 
     @abstractmethod
diff --git a/vllm/pooling_params.py b/vllm/pooling_params.py
index 5280e997c9..272552d61b 100644
--- a/vllm/pooling_params.py
+++ b/vllm/pooling_params.py
@@ -88,7 +88,7 @@ class PoolingParams(
         return deepcopy(self)
 
     def verify(self, model_config: ModelConfig) -> None:
-        # plugin task uses io_processor.parse_request to verify inputs,
+        # plugin task uses io_processor.parse_data to verify inputs,
         # skipping PoolingParams verify
         if self.task == "plugin":
             if self.skip_reading_prefix_cache is None:
diff --git a/vllm/v1/engine/async_llm.py b/vllm/v1/engine/async_llm.py
index 540ad23606..dd5cb2c262 100644
--- a/vllm/v1/engine/async_llm.py
+++ b/vllm/v1/engine/async_llm.py
@@ -339,8 +339,9 @@ class AsyncLLM(EngineClient):
         if isinstance(prompt, EngineCoreRequest):
             logger.warning_once(
                 "Passing EngineCoreRequest to AsyncLLM.generate() and .add_requests() "
-                "is deprecated and will be removed in v0.18. You should instead pass "
-                "the outputs of Renderer.render_cmpl() or Renderer.render_chat()."
+                "is deprecated and will be removed in the future. You should "
+                "instead pass the outputs of Renderer.render_cmpl() or "
+                "Renderer.render_chat()."
             )
 
             request = prompt
diff --git a/vllm/v1/engine/llm_engine.py b/vllm/v1/engine/llm_engine.py
index 274da62918..0db83b95f2 100644
--- a/vllm/v1/engine/llm_engine.py
+++ b/vllm/v1/engine/llm_engine.py
@@ -236,8 +236,9 @@ class LLMEngine:
         if isinstance(prompt, EngineCoreRequest):
             logger.warning_once(
                 "Passing EngineCoreRequest to LLMEngine.generate() and .add_requests() "
-                "is deprecated and will be removed in v0.18. You should instead pass "
-                "the outputs of Renderer.render_cmpl() or Renderer.render_chat()."
+                "is deprecated and will be removed in the future. You should "
+                "instead pass the outputs of Renderer.render_cmpl() or "
+                "Renderer.render_chat()."
             )
 
             request = prompt