diff --git a/cognee/infrastructure/databases/vector/embeddings/LiteLLMEmbeddingEngine.py b/cognee/infrastructure/databases/vector/embeddings/LiteLLMEmbeddingEngine.py
index 7aeb39b59..d83a2bcee 100644
--- a/cognee/infrastructure/databases/vector/embeddings/LiteLLMEmbeddingEngine.py
+++ b/cognee/infrastructure/databases/vector/embeddings/LiteLLMEmbeddingEngine.py
@@ -50,7 +50,7 @@ class LiteLLMEmbeddingEngine(EmbeddingEngine):
     dimensions: int
     mock: bool
 
-    MAX_RETRIES = 5
+    mock: bool
 
     def __init__(
         self,
@@ -71,7 +71,6 @@ class LiteLLMEmbeddingEngine(EmbeddingEngine):
         self.dimensions = dimensions
         self.max_completion_tokens = max_completion_tokens
         self.tokenizer = self.get_tokenizer()
-        self.retry_count = 0
         self.batch_size = batch_size
 
         enable_mocking = os.getenv("MOCK_EMBEDDING", "false")
@@ -195,40 +194,51 @@ class LiteLLMEmbeddingEngine(EmbeddingEngine):
             The tokenizer instance compatible with the model.
         """
         logger.debug(f"Loading tokenizer for model {self.model}...")
-        # If model also contains provider information, extract only model information
-        model = self.model.split("/")[-1]
+        
+        # Determine the effective model name.
+        # If the provider is OpenAI-compatible but not OpenAI itself, we might have a prefix like "openai/".
+        # We strip known prefixes to get the actual model ID for HuggingFace.
+        effective_model = self.model
+        if "hosted_vllm/" in effective_model:
+            effective_model = effective_model.replace("hosted_vllm/", "")
+        if "openai/" in effective_model:
+            effective_model = effective_model.replace("openai/", "")
 
-        if "openai" in self.provider.lower():
-            tokenizer = TikTokenTokenizer(
-                model=model, max_completion_tokens=self.max_completion_tokens
+        provider_lower = self.provider.lower()
+
+        if "openai" in provider_lower:
+             # TikToken expects the full model string usually, or at least how it was passed. 
+             # But if it's "openai/gpt-4", split acts on self.model.
+             # Original code: model = self.model.split("/")[-1]
+            return TikTokenTokenizer(
+                model=self.model.split("/")[-1], 
+                max_completion_tokens=self.max_completion_tokens
             )
-        elif "gemini" in self.provider.lower():
-            # Since Gemini tokenization needs to send an API request to get the token count we will use TikToken to
-            # count tokens as we calculate tokens word by word
-            tokenizer = TikTokenTokenizer(
-                model=None, max_completion_tokens=self.max_completion_tokens
+        
+        elif "gemini" in provider_lower:
+            return TikTokenTokenizer(
+                model=None, 
+                max_completion_tokens=self.max_completion_tokens
             )
-            # Note: Gemini Tokenizer expects an LLM model as input and not the embedding model
-            # tokenizer = GeminiTokenizer(
-            #     llm_model=llm_model, max_completion_tokens=self.max_completion_tokens
-            # )
-        elif "mistral" in self.provider.lower():
-            tokenizer = MistralTokenizer(
-                model=model, max_completion_tokens=self.max_completion_tokens
+
+        elif "mistral" in provider_lower:
+            return MistralTokenizer(
+                model=self.model.split("/")[-1],
+                max_completion_tokens=self.max_completion_tokens
             )
+
         else:
             try:
-                tokenizer = HuggingFaceTokenizer(
-                    model=self.model.replace("hosted_vllm/", "").replace("openai/", ""),
+                # Use the effective model name (stripped of prefixes) for HuggingFace
+                return HuggingFaceTokenizer(
+                    model=effective_model,
                     max_completion_tokens=self.max_completion_tokens,
                 )
             except Exception as e:
                 logger.warning(f"Could not get tokenizer from HuggingFace due to: {e}")
                 logger.info("Switching to TikToken default tokenizer.")
-                tokenizer = TikTokenTokenizer(
-                    model=None, max_completion_tokens=self.max_completion_tokens
+                return TikTokenTokenizer(
+                    model=None, 
+                    max_completion_tokens=self.max_completion_tokens
                 )
 
-        logger.debug(f"Tokenizer loaded for model: {self.model}")
-        return tokenizer
-