diff --git a/ddtrace/llmobs/_integrations/utils.py b/ddtrace/llmobs/_integrations/utils.py
index 11a7a408f84..3fbfb68b7c4 100644
--- a/ddtrace/llmobs/_integrations/utils.py
+++ b/ddtrace/llmobs/_integrations/utils.py
@@ -1168,9 +1168,9 @@ def openai_construct_tool_call_from_streamed_chunk(stored_tool_calls, tool_call_
 def openai_construct_message_from_streamed_chunks(streamed_chunks: list[Any]) -> dict[str, Any]:
     """Constructs a chat completion message dictionary from streamed chunks.
     The resulting message dictionary is of form:
-    {"content": "...", "role": "...", "tool_calls": [...], "finish_reason": "..."}
+    {"content": "...", "role": "...", "reasoning_content": "...", "tool_calls": [...], "finish_reason": "..."}
     """
-    message: dict[str, Any] = {"content": "", "tool_calls": []}
+    message: dict[str, Any] = {"content": "", "reasoning_content": "", "tool_calls": []}
     for chunk in streamed_chunks:
         if _get_attr(chunk, "usage", None):
             message["usage"] = chunk.usage
@@ -1182,6 +1182,9 @@ def openai_construct_message_from_streamed_chunks(streamed_chunks: list[Any]) ->
             message["role"] = chunk.delta.role
         if _get_attr(chunk, "finish_reason", None) and not message.get("finish_reason"):
             message["finish_reason"] = chunk.finish_reason
+        chunk_reasoning = _get_attr(chunk.delta, "reasoning_content", "")
+        if chunk_reasoning:
+            message["reasoning_content"] += chunk_reasoning
         chunk_content = _get_attr(chunk.delta, "content", "")
         if chunk_content:
             message["content"] += chunk_content
@@ -1198,6 +1201,8 @@ def openai_construct_message_from_streamed_chunks(streamed_chunks: list[Any]) ->
         message["tool_calls"].sort(key=lambda x: x.get("index", 0))
     else:
         message.pop("tool_calls", None)
+    if not message["reasoning_content"]:
+        message.pop("reasoning_content", None)
     message["content"] = message["content"].strip()
     return message
 
diff --git a/releasenotes/notes/fix-llmobs-streamed-reasoning-content-0da3242ccfaa6063.yaml b/releasenotes/notes/fix-llmobs-streamed-reasoning-content-0da3242ccfaa6063.yaml
new file mode 100644
index 00000000000..c8904ca689c
--- /dev/null
+++ b/releasenotes/notes/fix-llmobs-streamed-reasoning-content-0da3242ccfaa6063.yaml
@@ -0,0 +1,6 @@
+---
+fixes:
+  - |
+    LLM Observability: Fixes an issue where ``reasoning_content`` was missing from streamed chat completions
+    in the OpenAI and LiteLLM integrations when an OpenAI-compatible reasoning provider (e.g. DeepSeek, Qwen)
+    emitted ``delta.reasoning_content`` chunks. The aggregated message now captures reasoning text in the output message, matching non-streaming behavior.
diff --git a/tests/llmobs/test_integrations_utils.py b/tests/llmobs/test_integrations_utils.py
index eefd436b787..b29e475aae8 100644
--- a/tests/llmobs/test_integrations_utils.py
+++ b/tests/llmobs/test_integrations_utils.py
@@ -1,6 +1,9 @@
+from types import SimpleNamespace
+
 from ddtrace.llmobs._integrations.utils import _extract_chat_template_from_instructions
 from ddtrace.llmobs._integrations.utils import _normalize_prompt_variables
 from ddtrace.llmobs._integrations.utils import _openai_parse_input_response_messages
+from ddtrace.llmobs._integrations.utils import openai_construct_message_from_streamed_chunks
 
 
 def test_basic_functionality():
@@ -339,3 +342,51 @@ class FakeResponseReasoningItem:
         assert len(processed) == 1
         assert processed[0]["role"] == "user"
         assert tool_call_ids == []
+
+
+def _chunk(content=None, reasoning_content=None, role=None, finish_reason=None):
+    delta = SimpleNamespace(content=content, reasoning_content=reasoning_content, role=role)
+    return SimpleNamespace(delta=delta, finish_reason=finish_reason, usage=None, index=0)
+
+
+class TestOpenAIConstructMessageFromStreamedChunks:
+    def test_reasoning_then_content_chunks_aggregate_both(self):
+        # OpenAI-compatible reasoning providers (DeepSeek, Qwen, etc.) typically emit
+        # reasoning_content chunks first, then content chunks.
+        chunks = [
+            _chunk(role="assistant"),
+            _chunk(reasoning_content="Let me "),
+            _chunk(reasoning_content="think..."),
+            _chunk(content="The answer "),
+            _chunk(content="is 391."),
+            _chunk(finish_reason="stop"),
+        ]
+        message = openai_construct_message_from_streamed_chunks(chunks)
+        assert message["reasoning_content"] == "Let me think..."
+        assert message["content"] == "The answer is 391."
+        assert message["role"] == "assistant"
+        assert message["finish_reason"] == "stop"
+
+    def test_reasoning_only_stream(self):
+        chunks = [
+            _chunk(role="assistant"),
+            _chunk(reasoning_content="hmm"),
+        ]
+        message = openai_construct_message_from_streamed_chunks(chunks)
+        assert message["reasoning_content"] == "hmm"
+        assert message["content"] == ""
+
+    def test_no_reasoning_key_when_absent(self):
+        chunks = [_chunk(role="assistant"), _chunk(content="hello")]
+        message = openai_construct_message_from_streamed_chunks(chunks)
+        assert "reasoning_content" not in message
+        assert message["content"] == "hello"
+
+    def test_interleaved_reasoning_and_content_in_same_chunk(self):
+        chunks = [
+            _chunk(role="assistant"),
+            _chunk(reasoning_content="r", content="c"),
+        ]
+        message = openai_construct_message_from_streamed_chunks(chunks)
+        assert message["reasoning_content"] == "r"
+        assert message["content"] == "c"