diff --git a/python/packages/openai/agent_framework_openai/_chat_completion_client.py b/python/packages/openai/agent_framework_openai/_chat_completion_client.py
index 37c2e5b52dd..ad6a0dfef28 100644
--- a/python/packages/openai/agent_framework_openai/_chat_completion_client.py
+++ b/python/packages/openai/agent_framework_openai/_chat_completion_client.py
@@ -99,6 +99,54 @@
 ResponseModelT = TypeVar("ResponseModelT", bound=BaseModel | None, default=None)
 
 
+def _extract_reasoning_text(reasoning_details: Any) -> str | None:
+    """Extract visible plaintext from a reasoning_details payload.
+
+    OpenAI-compatible providers return reasoning_details in various forms:
+    - A list of dicts with ``"text"`` keys (e.g. OpenRouter: ``[{"type": "reasoning.text", "text": "..."}]``)
+    - A dict with a ``"content"`` key (list of text entries or a plain string)
+    - A plain string
+
+    Returns the concatenated plaintext if any is found, otherwise None.
+    """
+    if isinstance(reasoning_details, str):
+        return reasoning_details or None
+
+    if isinstance(reasoning_details, list):
+        parts: list[str] = []
+        for item in cast(list[object], reasoning_details):
+            if isinstance(item, str):
+                parts.append(item)
+            elif isinstance(item, dict):
+                entry_text = cast(dict[str, object], item).get("text")
+                if isinstance(entry_text, str) and entry_text:
+                    parts.append(entry_text)
+        return "".join(parts) or None
+
+    if isinstance(reasoning_details, dict):
+        detail_dict = cast(dict[str, object], reasoning_details)
+        # Handle {"text": "..."} style
+        text_val = detail_dict.get("text")
+        if isinstance(text_val, str) and text_val:
+            return text_val
+        # Handle {"content": [...]} or {"content": "..."} style
+        content_val = detail_dict.get("content")
+        if isinstance(content_val, str) and content_val:
+            return content_val
+        if isinstance(content_val, list):
+            parts = []
+            for item in cast(list[object], content_val):
+                if isinstance(item, str):
+                    parts.append(item)
+                elif isinstance(item, dict):
+                    entry_text2 = cast(dict[str, object], item).get("text")
+                    if isinstance(entry_text2, str) and entry_text2:
+                        parts.append(entry_text2)
+            return "".join(parts) or None
+
+    return None
+
+
 # region OpenAI Chat Options TypedDict
 
 
@@ -709,12 +757,14 @@ def _parse_response_from_openai(self, response: ChatCompletion, options: Mapping
             if choice.finish_reason:
                 finish_reason = choice.finish_reason  # type: ignore[assignment]
             contents: list[Content] = []
-            if text_content := self._parse_text_from_openai(choice):
-                contents.append(text_content)
+            contents.extend(self._parse_text_from_openai(choice))
             if parsed_tool_calls := [tool for tool in self._parse_tool_calls_from_openai(choice)]:
                 contents.extend(parsed_tool_calls)
             if reasoning_details := getattr(choice.message, "reasoning_details", None):
-                contents.append(Content.from_text_reasoning(protected_data=json.dumps(reasoning_details)))
+                contents.append(Content.from_text_reasoning(
+                    text=_extract_reasoning_text(reasoning_details),
+                    protected_data=json.dumps(reasoning_details),
+                ))
             messages.append(Message(role="assistant", contents=contents))
         return ChatResponse(
             response_id=response.id,
@@ -755,10 +805,12 @@ def _parse_response_update_from_openai(
                 continue
 
             contents.extend(self._parse_tool_calls_from_openai(choice))
-            if text_content := self._parse_text_from_openai(choice):
-                contents.append(text_content)
+            contents.extend(self._parse_text_from_openai(choice))
             if reasoning_details := getattr(choice.delta, "reasoning_details", None):
-                contents.append(Content.from_text_reasoning(protected_data=json.dumps(reasoning_details)))
+                contents.append(Content.from_text_reasoning(
+                    text=_extract_reasoning_text(reasoning_details),
+                    protected_data=json.dumps(reasoning_details),
+                ))
         return ChatResponseUpdate(
             created_at=datetime.fromtimestamp(chunk.created, tz=timezone.utc).strftime("%Y-%m-%dT%H:%M:%S.%fZ"),
             contents=contents,
@@ -795,14 +847,57 @@ def _parse_usage_from_openai(self, usage: CompletionUsage) -> UsageDetails:
                 details["cache_read_input_token_count"] = tokens
         return details
 
-    def _parse_text_from_openai(self, choice: Choice | ChunkChoice) -> Content | None:
-        """Parse the choice into a Content object with type='text'."""
+    def _parse_text_from_openai(self, choice: Choice | ChunkChoice) -> list[Content]:
+        """Parse the choice into Content objects with type='text' and/or 'text_reasoning'.
+
+        Handles both plain string content and structured list content (e.g. Mistral
+        reasoning models that return ``[{"type": "thinking", ...}, {"type": "text", ...}]``).
+        """
         message = choice.message if isinstance(choice, Choice) else choice.delta
         if message.content:
-            return Content.from_text(text=message.content, raw_representation=choice)
+            content = message.content
+            # Some OpenAI-compatible providers (e.g. Mistral reasoning models) return
+            # content as a list of typed chunks rather than a plain string.
+            if isinstance(content, list):
+                return self._parse_chunked_content(content, choice)
+            return [Content.from_text(text=content, raw_representation=choice)]
         if hasattr(message, "refusal") and message.refusal:
-            return Content.from_text(text=message.refusal, raw_representation=choice)
-        return None
+            return [Content.from_text(text=message.refusal, raw_representation=choice)]
+        return []
+
+    @staticmethod
+    def _parse_chunked_content(chunks: list[Any], choice: Choice | ChunkChoice) -> list[Content]:
+        """Parse structured content chunks (e.g. Mistral thinking/text format).
+
+        Handles chunk types:
+        - ``{"type": "thinking", "thinking": "..." | [{"type": "text", "text": "..."}]}``
+        - ``{"type": "text", "text": "..."}``
+        """
+        results: list[Content] = []
+        for item in cast(list[object], chunks):
+            if not isinstance(item, dict):
+                continue
+            chunk_dict = cast(dict[str, object], item)
+            chunk_type = chunk_dict.get("type")
+            if chunk_type == "thinking":
+                thinking = chunk_dict.get("thinking")
+                if isinstance(thinking, str):
+                    text = thinking
+                elif isinstance(thinking, list):
+                    text = "".join(
+                        str(cast(dict[str, object], part).get("text", ""))
+                        for part in cast(list[object], thinking)
+                        if isinstance(part, dict)
+                    )
+                else:
+                    text = ""
+                if text:
+                    results.append(Content.from_text_reasoning(text=text, raw_representation=choice))
+            elif chunk_type == "text":
+                text_val = chunk_dict.get("text")
+                if isinstance(text_val, str) and text_val:
+                    results.append(Content.from_text(text=text_val, raw_representation=choice))
+        return results
 
     def _get_metadata_from_chat_response(self, response: ChatCompletion) -> dict[str, Any]:
         """Get metadata from a chat response."""
diff --git a/python/packages/openai/tests/openai/test_openai_chat_completion_client.py b/python/packages/openai/tests/openai/test_openai_chat_completion_client.py
index 81d9963688b..30ac194913c 100644
--- a/python/packages/openai/tests/openai/test_openai_chat_completion_client.py
+++ b/python/packages/openai/tests/openai/test_openai_chat_completion_client.py
@@ -2087,4 +2087,222 @@ def test_streaming_chunk_with_null_delta_no_tool_calls_parsed(
     assert not any(c.type == "function_call" for c in update.contents)
 
 
+def test_parse_reasoning_details_extracts_plaintext(
+    openai_unit_test_env: dict[str, str],
+) -> None:
+    """Test that reasoning_details with plaintext entries surface text in Content.text (issue #6979)."""
+    from openai.types.chat.chat_completion import ChatCompletion, Choice
+    from openai.types.chat.chat_completion_message import ChatCompletionMessage
+
+    client = OpenAIChatCompletionClient()
+
+    # OpenRouter-style reasoning_details with plaintext
+    mock_reasoning_details = [
+        {"type": "reasoning.text", "text": "Let me think about this..."},
+        {"type": "reasoning.text", "text": " The answer is 42."},
+    ]
+
+    mock_response = ChatCompletion(
+        id="test-response",
+        object="chat.completion",
+        created=1234567890,
+        model="z-ai/glm-5.2",
+        choices=[
+            Choice(
+                index=0,
+                message=cast(Any, ChatCompletionMessage)(
+                    role="assistant",
+                    content="The answer is 42.",
+                    reasoning_details=mock_reasoning_details,
+                ),
+                finish_reason="stop",
+            )
+        ],
+    )
+
+    response = client._parse_response_from_openai(mock_response, {})
+
+    message = response.messages[0]
+    reasoning_contents = [c for c in message.contents if c.type == "text_reasoning"]
+    assert len(reasoning_contents) == 1
+
+    # Text should be extracted from the plaintext entries
+    assert reasoning_contents[0].text == "Let me think about this... The answer is 42."
+    # protected_data should still be preserved for round-trip
+    assert reasoning_contents[0].protected_data is not None
+    parsed = json.loads(reasoning_contents[0].protected_data)
+    assert parsed == mock_reasoning_details
+
+
+def test_parse_reasoning_details_extracts_plaintext_streaming(
+    openai_unit_test_env: dict[str, str],
+) -> None:
+    """Test streaming path extracts plaintext from reasoning_details (issue #6979)."""
+    from openai.types.chat.chat_completion_chunk import ChatCompletionChunk
+    from openai.types.chat.chat_completion_chunk import Choice as ChunkChoice
+    from openai.types.chat.chat_completion_chunk import ChoiceDelta as ChunkChoiceDelta
+
+    client = OpenAIChatCompletionClient()
+
+    mock_reasoning_details = [
+        {"type": "reasoning.text", "text": "Thinking step 1"},
+    ]
+
+    mock_chunk = ChatCompletionChunk(
+        id="test-chunk",
+        object="chat.completion.chunk",
+        created=1234567890,
+        model="z-ai/glm-5.2",
+        choices=[
+            ChunkChoice(
+                index=0,
+                delta=cast(Any, ChunkChoiceDelta)(
+                    role="assistant",
+                    content=None,
+                    reasoning_details=mock_reasoning_details,
+                ),
+                finish_reason=None,
+            )
+        ],
+    )
+
+    update = client._parse_response_update_from_openai(mock_chunk)
+
+    reasoning_contents = [c for c in update.contents if c.type == "text_reasoning"]
+    assert len(reasoning_contents) == 1
+    assert reasoning_contents[0].text == "Thinking step 1"
+    assert reasoning_contents[0].protected_data is not None
+
+
+def test_parse_mistral_chunked_content_from_response(
+    openai_unit_test_env: dict[str, str],
+) -> None:
+    """Test that Mistral-style list content (thinking + text chunks) is parsed correctly (issue #6978)."""
+    from openai.types.chat.chat_completion import ChatCompletion, Choice
+    from openai.types.chat.chat_completion_message import ChatCompletionMessage
+
+    client = OpenAIChatCompletionClient()
+
+    # Mistral returns content as a list of typed chunks.
+    # The OpenAI SDK passes this through without strict validation in streaming,
+    # so we use model_construct to bypass Pydantic validation in the test.
+    chunked_content = [
+        {"type": "thinking", "thinking": [{"type": "text", "text": "Let me reason about this..."}]},
+        {"type": "text", "text": "The answer is 42."},
+    ]
+
+    message = ChatCompletionMessage.model_construct(
+        role="assistant",
+        content=cast(Any, chunked_content),
+    )
+
+    mock_response = ChatCompletion(
+        id="test-response",
+        object="chat.completion",
+        created=1234567890,
+        model="mistral-medium-latest",
+        choices=[
+            Choice(
+                index=0,
+                message=message,
+                finish_reason="stop",
+            )
+        ],
+    )
+
+    response = client._parse_response_from_openai(mock_response, {})
+
+    message_out = response.messages[0]
+    assert len(message_out.contents) == 2
+
+    # First content should be text_reasoning from the thinking chunk
+    assert message_out.contents[0].type == "text_reasoning"
+    assert message_out.contents[0].text == "Let me reason about this..."
+
+    # Second content should be text from the text chunk
+    assert message_out.contents[1].type == "text"
+    assert message_out.contents[1].text == "The answer is 42."
+
+
+def test_parse_mistral_chunked_content_streaming(
+    openai_unit_test_env: dict[str, str],
+) -> None:
+    """Test streaming path handles Mistral-style list content (issue #6978)."""
+    from openai.types.chat.chat_completion_chunk import ChatCompletionChunk
+    from openai.types.chat.chat_completion_chunk import Choice as ChunkChoice
+    from openai.types.chat.chat_completion_chunk import ChoiceDelta as ChunkChoiceDelta
+
+    client = OpenAIChatCompletionClient()
+
+    # Streaming chunk with list content - use model_construct to bypass validation
+    chunked_content = [
+        {"type": "thinking", "thinking": "Deep thought..."},
+        {"type": "text", "text": "Here's the answer."},
+    ]
+
+    delta = ChunkChoiceDelta.model_construct(
+        role="assistant",
+        content=cast(Any, chunked_content),
+    )
+
+    mock_chunk = ChatCompletionChunk(
+        id="test-chunk",
+        object="chat.completion.chunk",
+        created=1234567890,
+        model="mistral-medium-latest",
+        choices=[
+            ChunkChoice(
+                index=0,
+                delta=delta,
+                finish_reason=None,
+            )
+        ],
+    )
+
+    update = client._parse_response_update_from_openai(mock_chunk)
+
+    reasoning_contents = [c for c in update.contents if c.type == "text_reasoning"]
+    text_contents = [c for c in update.contents if c.type == "text"]
+
+    assert len(reasoning_contents) == 1
+    assert reasoning_contents[0].text == "Deep thought..."
+
+    assert len(text_contents) == 1
+    assert text_contents[0].text == "Here's the answer."
+
+
+def test_parse_plain_string_content_still_works(
+    openai_unit_test_env: dict[str, str],
+) -> None:
+    """Regression test: plain string content still works after list-content handling."""
+    from openai.types.chat.chat_completion import ChatCompletion, Choice
+    from openai.types.chat.chat_completion_message import ChatCompletionMessage
+
+    client = OpenAIChatCompletionClient()
+
+    mock_response = ChatCompletion(
+        id="test-response",
+        object="chat.completion",
+        created=1234567890,
+        model="gpt-4o",
+        choices=[
+            Choice(
+                index=0,
+                message=cast(Any, ChatCompletionMessage)(
+                    role="assistant",
+                    content="Just a normal string response.",
+                ),
+                finish_reason="stop",
+            )
+        ],
+    )
+
+    response = client._parse_response_from_openai(mock_response, {})
+
+    message = response.messages[0]
+    assert len(message.contents) == 1
+    assert message.contents[0].type == "text"
+    assert message.contents[0].text == "Just a normal string response."
+
+
 # endregion