diff --git a/config.py b/config.py
index df7a987..4ef22f9 100644
--- a/config.py
+++ b/config.py
@@ -61,7 +61,7 @@ os.environ["LLAMA_CLOUD_API_KEY"] = LLAMA_CLOUD_API_KEY
 print(f"OpenAI API key {'is set' if OPENAI_API_KEY else 'is NOT set'}", file=sys.stderr)
 
 # --- AI Model Configuration ---
-LLM_MODEL = "chatgpt-4o-latest" # Or "gpt-4o" etc.
+LLM_MODEL = "gpt-5.2" # Or "gpt-4o" etc.
 EMBEDDING_MODEL = "text-embedding-3-small"
 LLM_TEMPERATURE = 0.3
 LLM_TIMEOUT = 300.0  # 5 minutes
diff --git a/graph_rag_integration.py b/graph_rag_integration.py
index a0dd2e0..6204899 100644
--- a/graph_rag_integration.py
+++ b/graph_rag_integration.py
@@ -219,7 +219,7 @@ class GraphRAGStore:
             for i, chunk in enumerate(chunks):
                 try:
                     # Use GPT-4o-mini model for better cost efficiency
-                    llm = OpenAI(model="gpt-4o-mini")
+                    llm = OpenAI(model="gpt-5-mini")
                     messages = [
                         ChatMessage(
                             role="system",
@@ -237,7 +237,7 @@ class GraphRAGStore:
             if summaries:
                 final_summary_text = "\n\n".join(summaries)
                 try:
-                    llm = OpenAI(model="gpt-4o-mini")
+                    llm = OpenAI(model="gpt-5-mini")
                     messages = [
                         ChatMessage(
                             role="system",
@@ -257,7 +257,7 @@ class GraphRAGStore:
         # For normal size text, use the larger model directly
         try:
             # Use GPT-4o-mini model for better cost efficiency
-            llm = OpenAI(model="gpt-4o-mini")
+            llm = OpenAI(model="gpt-5-mini")
             messages = [
                 ChatMessage(
                     role="system",
@@ -825,9 +825,9 @@ def generate_final_answer(query, retrieval_result, llm):
     
     # If no model was provided or we're forcing to use a specific model
     if llm is None or not hasattr(llm, 'chat'):
-        # Fallback to gpt-4o-mini for better cost efficiency
-        llm = OpenAI(model="gpt-4o-mini")
-        log_structured('info', 'Using gpt-4o-mini model for final answer generation')
+        # Fallback to gpt-5-mini for better cost efficiency
+        llm = OpenAI(model="gpt-5-mini")
+        log_structured('info', 'Using gpt-5-mini model for final answer generation')
     
     prompt = f"""
     Based on the following information from two different sources, please answer this question: {query}
diff --git a/mongodb_utils.py b/mongodb_utils.py
index 75086be..e138ab5 100644
--- a/mongodb_utils.py
+++ b/mongodb_utils.py
@@ -327,7 +327,7 @@ def generate_conversation_title(conversation_id: str, content: List[Dict]) -> Op
         
         # Create LLM instance
         llm = LlamaOpenAI(
-            model="chatgpt-4o-latest",
+            model="gpt-5.2",
             temperature=0.3,
         )