diff --git a/src/strands/models/openai_responses.py b/src/strands/models/openai_responses.py
index bc2dcfd0e..b2f93ae72 100644
--- a/src/strands/models/openai_responses.py
+++ b/src/strands/models/openai_responses.py
@@ -469,7 +469,7 @@ def _format_request_messages(cls, messages: Messages) -> list[dict[str, Any]]:
             formatted_contents = [
                 cls._format_request_message_content(content, role=role)
                 for content in contents
-                if not any(block_type in content for block_type in ["toolResult", "toolUse"])
+                if not any(block_type in content for block_type in ["toolResult", "toolUse", "reasoningContent"])
             ]
 
             formatted_tool_calls = [
diff --git a/src/strands/models/writer.py b/src/strands/models/writer.py
index 94774b363..0e774134f 100644
--- a/src/strands/models/writer.py
+++ b/src/strands/models/writer.py
@@ -116,7 +116,7 @@ def _format_content_vision(content: ContentBlock) -> dict[str, Any]:
         return [
             _format_content_vision(content)
             for content in contents
-            if not any(block_type in content for block_type in ["toolResult", "toolUse"])
+            if not any(block_type in content for block_type in ["toolResult", "toolUse", "reasoningContent"])
         ]
 
     def _format_request_message_contents(self, contents: list[ContentBlock]) -> str:
@@ -142,7 +142,7 @@ def _format_content(content: ContentBlock) -> str:
         content_blocks = list(
             filter(
                 lambda content: content.get("text")
-                and not any(block_type in content for block_type in ["toolResult", "toolUse"]),
+                and not any(block_type in content for block_type in ["toolResult", "toolUse", "reasoningContent"]),
                 contents,
             )
         )
diff --git a/tests/strands/models/test_openai_responses.py b/tests/strands/models/test_openai_responses.py
index 545f128bf..012383f6a 100644
--- a/tests/strands/models/test_openai_responses.py
+++ b/tests/strands/models/test_openai_responses.py
@@ -317,6 +317,87 @@ def test_format_request_message_content_role_user():
     assert result == {"type": "input_text", "text": "question"}
 
 
+def test_format_request_messages_strips_reasoning_content():
+    """reasoningContent blocks from Gemini are silently dropped when formatting for OpenAI.
+
+    In multi-provider Swarms a Gemini agent may store reasoningContent blocks in the shared
+    conversation history. These blocks have no OpenAI equivalent and must be filtered out
+    before the request is sent, otherwise _format_request_message_content raises TypeError.
+
+    Regression test for https://github.com/strands-agents/sdk-python/issues/1993
+    """
+    messages = [
+        {
+            "role": "user",
+            "content": [{"text": "What is AI?"}],
+        },
+        {
+            # Simulates an assistant message produced by a Gemini agent in a Swarm —
+            # contains both a reasoningContent block and the visible text response.
+            "role": "assistant",
+            "content": [
+                {
+                    "reasoningContent": {
+                        "reasoningText": {
+                            "text": "The user asked about AI. I should give a brief answer.",
+                            "signature": "SGVsbG8gV29ybGQ=",  # base64 blob Gemini attaches
+                        }
+                    }
+                },
+                {"text": "AI is the simulation of human intelligence by machines."},
+            ],
+        },
+    ]
+
+    # Must not raise TypeError
+    result = OpenAIResponsesModel._format_request_messages(messages)
+
+    # The reasoningContent block is dropped; only the visible text survives
+    assert result == [
+        {
+            "role": "user",
+            "content": [{"type": "input_text", "text": "What is AI?"}],
+        },
+        {
+            "role": "assistant",
+            "content": [
+                {"type": "output_text", "text": "AI is the simulation of human intelligence by machines."}
+            ],
+        },
+    ]
+
+
+def test_format_request_messages_reasoning_content_only_message_is_dropped():
+    """An assistant message consisting solely of a reasoningContent block is removed entirely.
+
+    When Gemini emits a turn that is pure reasoning (no visible text), the resulting
+    formatted message would have an empty content list and should be omitted.
+    """
+    messages = [
+        {"role": "user", "content": [{"text": "Think first."}]},
+        {
+            "role": "assistant",
+            "content": [
+                {
+                    "reasoningContent": {
+                        "reasoningText": {"text": "Let me think...", "signature": ""}
+                    }
+                }
+            ],
+        },
+        {"role": "user", "content": [{"text": "Now answer."}]},
+    ]
+
+    result = OpenAIResponsesModel._format_request_messages(messages)
+
+    # The pure-reasoning assistant message is dropped entirely (empty content filtered at line 498-502)
+    roles = [item.get("role") for item in result if "role" in item]
+    assert roles == ["user", "user"]
+    assert all(
+        "reasoningContent" not in str(item) for item in result
+    )
+
+
 def test_format_request(model, messages, tool_specs, system_prompt):
     tru_request = model._format_request(messages, tool_specs, system_prompt)
     exp_request = {
diff --git a/tests/strands/models/test_writer.py b/tests/strands/models/test_writer.py
index 81745f412..1fa6d9ff8 100644
--- a/tests/strands/models/test_writer.py
+++ b/tests/strands/models/test_writer.py
@@ -250,7 +250,6 @@ def test_format_request_with_empty_content(model, model_id, stream_options):
     [
         ({"video": {}}, "video"),
         ({"document": {}}, "document"),
-        ({"reasoningContent": {}}, "reasoningContent"),
         ({"other": {}}, "other"),
     ],
 )
@@ -266,6 +265,44 @@ def test_format_request_with_unsupported_type(model, content, content_type):
         model.format_request(messages)
 
 
+def test_format_request_strips_reasoning_content(model, model_id, stream_options):
+    """reasoningContent blocks from Gemini are silently dropped when formatting for Writer.
+
+    In multi-provider Swarms a Gemini agent may store reasoningContent blocks in the shared
+    conversation history. These blocks have no Writer equivalent and must be filtered out
+    before the request is sent, otherwise _format_request_message_content raises TypeError.
+
+    Regression test for https://github.com/strands-agents/sdk-python/issues/1993
+    """
+    messages = [
+        {"role": "user", "content": [{"text": "What is AI?"}]},
+        {
+            "role": "assistant",
+            "content": [
+                {
+                    "reasoningContent": {
+                        "reasoningText": {
+                            "text": "The user asked about AI.",
+                            "signature": "SGVsbG8gV29ybGQ=",
+                        }
+                    }
+                },
+                {"text": "AI is the simulation of human intelligence by machines."},
+            ],
+        },
+    ]
+
+    # Must not raise TypeError
+    tru_request = model.format_request(messages)
+
+    # The reasoningContent block is dropped; only the visible text message survives
+    assistant_messages = [m for m in tru_request["messages"] if m.get("role") == "assistant"]
+    assert len(assistant_messages) == 1
+    assert assistant_messages[0]["content"] == [
+        {"type": "text", "text": "AI is the simulation of human intelligence by machines."}
+    ]
+
+
 class AsyncStreamWrapper:
     def __init__(self, items: list[Any]):
         self.items = items