diff --git a/src/strands/models/openai.py b/src/strands/models/openai.py
index 73484e924..07c019a90 100644
--- a/src/strands/models/openai.py
+++ b/src/strands/models/openai.py
@@ -440,6 +440,12 @@ def format_request_messages(
 
         return [message for message in formatted_messages if "content" in message or "tool_calls" in message]
 
+    def _default_stop_tokens(self) -> list[str] | None:
+        model_id = str(self.config.get("model_id", "")).lower()
+        if "gpt-oss" in model_id:
+            return ["<|call|>", "<|return|>", "<|end|>"]
+        return None
+
     def format_request(
         self,
         messages: Messages,
@@ -467,6 +473,12 @@ def format_request(
             TypeError: If a message contains a content block type that cannot be converted to an OpenAI-compatible
                 format.
         """
+        params = dict(cast(dict[str, Any], self.config.get("params", {})))
+        if "stop" not in params:
+            default_stop = self._default_stop_tokens()
+            if default_stop:
+                params["stop"] = default_stop
+
         return {
             "messages": self.format_request_messages(
                 messages, system_prompt, system_prompt_content=system_prompt_content
@@ -486,7 +498,7 @@ def format_request(
                 for tool_spec in tool_specs or []
             ],
             **(self._format_request_tool_choice(tool_choice)),
-            **cast(dict[str, Any], self.config.get("params", {})),
+            **params,
         }
 
     def format_chunk(self, event: dict[str, Any], **kwargs: Any) -> StreamEvent:
diff --git a/tests/strands/models/test_openai.py b/tests/strands/models/test_openai.py
index 747e1123a..1a338371a 100644
--- a/tests/strands/models/test_openai.py
+++ b/tests/strands/models/test_openai.py
@@ -627,6 +627,25 @@ def test_format_request(model, messages, tool_specs, system_prompt):
     assert tru_request == exp_request
 
 
+def test_format_request_adds_gpt_oss_stop_tokens(messages, tool_specs, system_prompt):
+    model = OpenAIModel(model_id="openai/gpt-oss-120b", params={"max_tokens": 1})
+
+    tru_request = model.format_request(messages, tool_specs, system_prompt)
+
+    assert tru_request["stop"] == ["<|call|>", "<|return|>", "<|end|>"]
+
+
+def test_format_request_preserves_explicit_stop_tokens(messages, tool_specs, system_prompt):
+    model = OpenAIModel(
+        model_id="openai/gpt-oss-120b",
+        params={"max_tokens": 1, "stop": ["<|end|>"]},
+    )
+
+    tru_request = model.format_request(messages, tool_specs, system_prompt)
+
+    assert tru_request["stop"] == ["<|end|>"]
+
+
 def test_format_request_with_tool_choice_auto(model, messages, tool_specs, system_prompt):
     tool_choice = {"auto": {}}
     tru_request = model.format_request(messages, tool_specs, system_prompt, tool_choice)