diff --git a/optimum/intel/openvino/modeling_base.py b/optimum/intel/openvino/modeling_base.py
index 99422f1a54..b32ee4908b 100644
--- a/optimum/intel/openvino/modeling_base.py
+++ b/optimum/intel/openvino/modeling_base.py
@@ -592,7 +592,7 @@ def _from_transformers(
         if load_in_8bit is None and not quantization_config:
             ov_config = None
         else:
-            ov_config = OVConfig(dtype="fp32")
+            ov_config = OVConfig(dtype="auto")
 
         main_export(
             model_name_or_path=model_id,
diff --git a/optimum/intel/openvino/modeling_base_seq2seq.py b/optimum/intel/openvino/modeling_base_seq2seq.py
index d01c396a42..69c4bfab10 100644
--- a/optimum/intel/openvino/modeling_base_seq2seq.py
+++ b/optimum/intel/openvino/modeling_base_seq2seq.py
@@ -364,7 +364,7 @@ def _from_transformers(
         if load_in_8bit is None and not quantization_config:
             ov_config = None
         else:
-            ov_config = OVConfig(dtype="fp32")
+            ov_config = OVConfig(dtype="auto")
 
         main_export(
             model_name_or_path=model_id,
diff --git a/optimum/intel/openvino/modeling_open_clip.py b/optimum/intel/openvino/modeling_open_clip.py
index ef00c182e5..0df51cfd69 100644
--- a/optimum/intel/openvino/modeling_open_clip.py
+++ b/optimum/intel/openvino/modeling_open_clip.py
@@ -247,7 +247,7 @@ def _from_transformers(
         if load_in_8bit is None and not quantization_config:
             ov_config = None
         else:
-            ov_config = OVConfig(dtype="fp32")
+            ov_config = OVConfig(dtype="auto")
 
         def fn_get_submodels(model):
             return {"model_text": model.text}
@@ -372,7 +372,7 @@ def _from_transformers(
         if load_in_8bit is None and not quantization_config:
             ov_config = None
         else:
-            ov_config = OVConfig(dtype="fp32")
+            ov_config = OVConfig(dtype="auto")
 
         def fn_get_submodels(model):
             return {"model_vision": model.visual}
diff --git a/optimum/intel/openvino/modeling_visual_language.py b/optimum/intel/openvino/modeling_visual_language.py
index 8d6edea0f0..02ceab177d 100644
--- a/optimum/intel/openvino/modeling_visual_language.py
+++ b/optimum/intel/openvino/modeling_visual_language.py
@@ -598,7 +598,7 @@ def _from_transformers(
         if load_in_8bit is None and not quantization_config:
             ov_config = None
         else:
-            ov_config = OVConfig(dtype="fp32")
+            ov_config = OVConfig(dtype="auto")
 
         stateful = kwargs.pop("stateful", ensure_stateful_is_available(warn=False) and use_cache)