diff --git a/src/transformers/__init__.py b/src/transformers/__init__.py
index 96937344cd8099..39cb0eabe67860 100644
--- a/src/transformers/__init__.py
+++ b/src/transformers/__init__.py
@@ -773,7 +773,6 @@
         "SiglipConfig",
         "SiglipProcessor",
         "SiglipTextConfig",
-        "SiglipTokenizer",
         "SiglipVisionConfig",
     ],
     "models.speech_encoder_decoder": ["SpeechEncoderDecoderConfig"],
@@ -1125,6 +1124,7 @@
     _import_structure["models.reformer"].append("ReformerTokenizer")
     _import_structure["models.rembert"].append("RemBertTokenizer")
     _import_structure["models.seamless_m4t"].append("SeamlessM4TTokenizer")
+    _import_structure["models.siglip"].append("SiglipTokenizer")
     _import_structure["models.speech_to_text"].append("Speech2TextTokenizer")
     _import_structure["models.speecht5"].append("SpeechT5Tokenizer")
     _import_structure["models.t5"].append("T5Tokenizer")
@@ -5503,7 +5503,6 @@
         SiglipConfig,
         SiglipProcessor,
         SiglipTextConfig,
-        SiglipTokenizer,
         SiglipVisionConfig,
     )
     from .models.speech_encoder_decoder import SpeechEncoderDecoderConfig
@@ -5852,6 +5851,7 @@
         from .models.reformer import ReformerTokenizer
         from .models.rembert import RemBertTokenizer
         from .models.seamless_m4t import SeamlessM4TTokenizer
+        from .models.siglip import SiglipTokenizer
         from .models.speech_to_text import Speech2TextTokenizer
         from .models.speecht5 import SpeechT5Tokenizer
         from .models.t5 import T5Tokenizer
diff --git a/src/transformers/models/siglip/__init__.py b/src/transformers/models/siglip/__init__.py
index d47b7c5f5d02f6..f802f630af7867 100644
--- a/src/transformers/models/siglip/__init__.py
+++ b/src/transformers/models/siglip/__init__.py
@@ -16,6 +16,7 @@
 from ...utils import (
     OptionalDependencyNotAvailable,
     _LazyModule,
+    is_sentencepiece_available,
     is_torch_available,
     is_vision_available,
 )
@@ -29,9 +30,17 @@
         "SiglipVisionConfig",
     ],
     "processing_siglip": ["SiglipProcessor"],
-    "tokenization_siglip": ["SiglipTokenizer"],
 }
 
+try:
+    if not is_sentencepiece_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["tokenization_siglip"] = ["SiglipTokenizer"]
+
+
 try:
     if not is_vision_available():
         raise OptionalDependencyNotAvailable()
@@ -63,7 +72,14 @@
         SiglipVisionConfig,
     )
     from .processing_siglip import SiglipProcessor
-    from .tokenization_siglip import SiglipTokenizer
+
+    try:
+        if not is_sentencepiece_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .tokenization_siglip import SiglipTokenizer
 
     try:
         if not is_vision_available():
diff --git a/src/transformers/utils/dummy_sentencepiece_objects.py b/src/transformers/utils/dummy_sentencepiece_objects.py
index 658645746329c6..6064ef7e8cac3c 100644
--- a/src/transformers/utils/dummy_sentencepiece_objects.py
+++ b/src/transformers/utils/dummy_sentencepiece_objects.py
@@ -184,6 +184,13 @@ def __init__(self, *args, **kwargs):
         requires_backends(self, ["sentencepiece"])
 
 
+class SiglipTokenizer(metaclass=DummyObject):
+    _backends = ["sentencepiece"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["sentencepiece"])
+
+
 class Speech2TextTokenizer(metaclass=DummyObject):
     _backends = ["sentencepiece"]