diff --git a/vllm/renderers/deepseek_v32.py b/vllm/renderers/deepseek_v32.py
index 18103b9d7e5e..91b95db06f41 100644
--- a/vllm/renderers/deepseek_v32.py
+++ b/vllm/renderers/deepseek_v32.py
@@ -71,7 +71,6 @@ def render_messages(
             content_format="string",
         )
 
-        kwargs["return_dict"] = False
         prompt_raw = tokenizer.apply_chat_template(
             conversation=conversation,
             messages=messages,
@@ -98,7 +97,6 @@ async def render_messages_async(
             content_format="string",
         )
 
-        kwargs["return_dict"] = False
         prompt_raw = tokenizer.apply_chat_template(
             conversation=conversation,
             messages=messages,
diff --git a/vllm/renderers/grok2.py b/vllm/renderers/grok2.py
index e064b7c8bbdf..feefe8f0b4e8 100644
--- a/vllm/renderers/grok2.py
+++ b/vllm/renderers/grok2.py
@@ -71,7 +71,6 @@ def render_messages(
             content_format="string",
         )
 
-        kwargs["return_dict"] = False
         prompt_raw = tokenizer.apply_chat_template(
             conversation=conversation,
             messages=messages,
@@ -98,7 +97,6 @@ async def render_messages_async(
             content_format="string",
         )
 
-        kwargs["return_dict"] = False
         prompt_raw = tokenizer.apply_chat_template(
             conversation=conversation,
             messages=messages,
diff --git a/vllm/renderers/hf.py b/vllm/renderers/hf.py
index 41a8aa450c9f..eb90c9e6d3a6 100644
--- a/vllm/renderers/hf.py
+++ b/vllm/renderers/hf.py
@@ -466,7 +466,6 @@ def safe_apply_chat_template(
         chat_template=chat_template,
         chat_template_kwargs=kwargs,
     )
-    resolved_kwargs["return_dict"] = False
 
     try:
         return tokenizer.apply_chat_template(
diff --git a/vllm/renderers/params.py b/vllm/renderers/params.py
index 19555bba04b1..d20038478ded 100644
--- a/vllm/renderers/params.py
+++ b/vllm/renderers/params.py
@@ -66,7 +66,7 @@ def get_apply_chat_template_kwargs(self) -> dict[str, Any]:
         """The arguments to pass to `tokenizer.apply_chat_template`."""
         return merge_kwargs(
             self.chat_template_kwargs,
-            dict(chat_template=self.chat_template),
+            dict(chat_template=self.chat_template, return_dict=False),
         )