diff --git a/utils/convert-hf-to-gguf-bitnet.py b/utils/convert-hf-to-gguf-bitnet.py
index 23e84384c..0a6b8c560 100644
--- a/utils/convert-hf-to-gguf-bitnet.py
+++ b/utils/convert-hf-to-gguf-bitnet.py
@@ -957,7 +957,7 @@ class BitnetModel(Model):
     model_arch = gguf.MODEL_ARCH.BITNET
 
     def set_vocab(self):
-        self._set_vocab_sentencepiece()
+        self._set_vocab_gpt2()
         
     def set_gguf_parameters(self):
         super().set_gguf_parameters()
diff --git a/utils/convert-ms-to-gguf-bitnet.py b/utils/convert-ms-to-gguf-bitnet.py
index edf702788..1b9b8d901 100644
--- a/utils/convert-ms-to-gguf-bitnet.py
+++ b/utils/convert-ms-to-gguf-bitnet.py
@@ -1208,6 +1208,9 @@ def extract_vocabulary_from_model(self, vocab: Vocab) -> tuple[list[bytes], list
     def add_meta_vocab(self, vocab: Vocab) -> None:
         # Ensure that tokenizer_model is added to the GGUF model
         self.gguf.add_tokenizer_model(vocab.tokenizer_model)
+        # Add pre-tokenizer type for BPE models (required for correct tokenization)
+        if vocab.tokenizer_model == "gpt2":
+            self.gguf.add_token_pre_type("gpt-2")
         # Extract model vocabulary for model conversion
         tokens, scores, toktypes = self.extract_vocabulary_from_model(vocab)