diff --git a/packages/graphrag/graphrag/index/operations/summarize_communities/community_reports_extractor.py b/packages/graphrag/graphrag/index/operations/summarize_communities/community_reports_extractor.py
index 4513abe95b..efae3cb82e 100644
--- a/packages/graphrag/graphrag/index/operations/summarize_communities/community_reports_extractor.py
+++ b/packages/graphrag/graphrag/index/operations/summarize_communities/community_reports_extractor.py
@@ -49,15 +49,35 @@ class CommunityReportsResult:
     structured_output: CommunityReportResponse | None
 
 
-class CommunityReportsExtractor:
-    """Community reports extractor class definition."""
+def _is_unsupported_response_format_error(e: Exception) -> bool:
+    """Check if the error is due to unsupported response_format."""
+    msg = str(e).lower()
+    return any(
+        phrase in msg
+        for phrase in [
+            "response_format",
+            "json_schema",
+            "unsupported",
+            "unavailable",
+        ]
+    )
+
 
-    _model: "LLMCompletion"
-    _extraction_prompt: str
-    _output_formatter_prompt: str
-    _on_error: ErrorHandlerFn
-    _max_report_length: int
+def _parse_json_from_text(
+    text: str, model: type[CommunityReportResponse]
+) -> CommunityReportResponse:
+    """Extract and parse JSON from LLM text output."""
+    text = text or ""
+    # Try code block first
+    m = re.search(r"```(?:json)?\s*(\{.*?\})\s*```", text, re.DOTALL)
+    if not m:
+        m = re.search(r"(\{.*\})", text, re.DOTALL)
+    json_text = m.group(1).strip() if m else text.strip()
+    data = json.loads(json_text)
+    return model(**data)
 
+
+class CommunityReportsExtractor:
     def __init__(
         self,
         model: "LLMCompletion",
@@ -65,26 +85,61 @@ def __init__(
         max_report_length: int,
         on_error: ErrorHandlerFn | None = None,
     ):
-        """Init method definition."""
         self._model = model
         self._extraction_prompt = extraction_prompt
         self._on_error = on_error or (lambda _e, _s, _d: None)
         self._max_report_length = max_report_length
 
     async def __call__(self, input_text: str):
-        """Call method definition."""
         output = None
         try:
             prompt = self._extraction_prompt.format(**{
                 INPUT_TEXT_KEY: input_text,
                 MAX_LENGTH_KEY: str(self._max_report_length),
             })
-            response = await self._model.completion_async(
-                messages=prompt,
-                response_format=CommunityReportResponse,  # A model is required when using json mode
-            )
 
-            output = response.formatted_response  # type: ignore
+            # Strategy 1: Try structured output with Pydantic model (json_schema)
+            try:
+                response = await self._model.completion_async(
+                    messages=prompt,
+                    response_format=CommunityReportResponse,
+                )
+                output = response.formatted_response
+            except Exception as schema_error:
+                if not _is_unsupported_response_format_error(schema_error):
+                    raise
+
+                logger.warning(
+                    "json_schema not supported by provider, "
+                    "falling back to json_object mode"
+                )
+
+                # Strategy 2: Fallback to json_object
+                try:
+                    response = await self._model.completion_async(
+                        messages=prompt,
+                        response_format={"type": "json_object"},
+                    )
+                    output = _parse_json_from_text(
+                        response.content, CommunityReportResponse
+                    )
+                except Exception as json_error:
+                    if not _is_unsupported_response_format_error(json_error):
+                        raise
+
+                    logger.warning(
+                        "json_object not supported by provider, "
+                        "falling back to plain text parsing"
+                    )
+
+                    # Strategy 3: Final fallback to plain text
+                    response = await self._model.completion_async(
+                        messages=prompt,
+                    )
+                    output = _parse_json_from_text(
+                        response.content, CommunityReportResponse
+                    )
+
         except Exception as e:
             logger.exception("error generating community report")
             self._on_error(e, traceback.format_exc(), None)