diff --git a/mteb/models/openai_models.py b/mteb/models/openai_models.py
index 8588104081..dce63fe9ea 100644
--- a/mteb/models/openai_models.py
+++ b/mteb/models/openai_models.py
@@ -18,9 +18,10 @@ class OpenAIWrapper(Wrapper):
     def __init__(
         self,
         model_name: str,
+        *,
         max_tokens: int,
+        embed_dim: int,
         tokenizer_name: str = "cl100k_base",  # since all models use this tokenizer now
-        embed_dim: int | None = None,
         **kwargs,
     ) -> None:
         """Wrapper for OpenAIs embedding API.
@@ -54,17 +55,23 @@ def truncate_text_tokens(self, text):
         return self._encoding.decode(truncated_sentence)
 
     def encode(self, sentences: list[str], **kwargs: Any) -> np.ndarray:
-        requires_package(self, "openai", "Openai text embedding")
-
+        print("Starting OpenAIWrapper encode method")
         from openai import NotGiven
 
+        empty_mask = [not s.strip() for s in sentences]
         if self._model_name == "text-embedding-ada-002" and self._embed_dim is not None:
             logger.warning(
                 "Reducing embedding size available only for text-embedding-3-* models"
             )
 
+        if any(empty_mask):
+            logger.warning(
+                "Empty strings detected - encoding non-empty only, returing zero vectors for empty strings."
+            )
+
+        non_empty_sentences = [s for s, e in zip(sentences, empty_mask) if not e]
         trimmed_sentences = []
-        for sentence in sentences:
+        for sentence in non_empty_sentences:
             encoded_sentence = self._encoding.encode(sentence)
             if len(encoded_sentence) > self._max_tokens:
                 truncated_sentence = self.truncate_text_tokens(sentence)
@@ -118,7 +125,11 @@ def encode(self, sentences: list[str], **kwargs: Any) -> np.ndarray:
                     )
             all_embeddings.extend(self._to_numpy(response))
 
-        return np.array(all_embeddings)
+        all_embeddings = np.array(all_embeddings)
+        final_embeddings = np.zeros((len(sentences), self._embed_dim), dtype=np.float32)
+        non_empty_idxs = [i for i, e in enumerate(empty_mask) if not e]
+        final_embeddings[non_empty_idxs] = all_embeddings
+        return final_embeddings
 
     def _to_numpy(self, embedding_response) -> np.ndarray:
         return np.array([e.embedding for e in embedding_response.data])
@@ -134,6 +145,7 @@ def _to_numpy(self, embedding_response) -> np.ndarray:
         model_name="text-embedding-3-small",
         tokenizer_name="cl100k_base",
         max_tokens=8191,
+        embed_dim=1536,
     ),
     max_tokens=8191,
     embed_dim=1536,
@@ -159,6 +171,7 @@ def _to_numpy(self, embedding_response) -> np.ndarray:
         model_name="text-embedding-3-large",
         tokenizer_name="cl100k_base",
         max_tokens=8191,
+        embed_dim=3072,
     ),
     max_tokens=8191,
     embed_dim=3072,
@@ -184,6 +197,7 @@ def _to_numpy(self, embedding_response) -> np.ndarray:
         model_name="text-embedding-ada-002",
         tokenizer_name="cl100k_base",
         max_tokens=8191,
+        embed_dim=1536,
     ),
     reference="https://openai.com/index/new-and-improved-embedding-model/",
     max_tokens=8191,