diff --git a/aphrodite/common/config.py b/aphrodite/common/config.py
index 4a490e4de..ce5e19afd 100644
--- a/aphrodite/common/config.py
+++ b/aphrodite/common/config.py
@@ -19,6 +19,7 @@
 from aphrodite.platforms import current_platform
 from aphrodite.quantization import QUANTIZATION_METHODS
 from aphrodite.transformers_utils.config import (ConfigFormat, get_config,
+                                                 get_hf_image_processor_config,
                                                  get_hf_text_config)
 from aphrodite.triton_utils import HAS_TRITON
 
@@ -203,6 +204,8 @@ def __init__(
                                     code_revision, rope_scaling, rope_theta,
                                     config_format)
         self.hf_text_config = get_hf_text_config(self.hf_config)
+        self.hf_image_processor_config = get_hf_image_processor_config(
+            self.model, revision)
         self.dtype = _get_and_verify_dtype(self.hf_text_config, dtype)
 
         # Choose a default enforce_eager value if the user did not specify
diff --git a/aphrodite/inputs/registry.py b/aphrodite/inputs/registry.py
index 5f81c7031..e567b7bea 100644
--- a/aphrodite/inputs/registry.py
+++ b/aphrodite/inputs/registry.py
@@ -2,8 +2,8 @@
 from array import array
 from collections import UserDict
 from dataclasses import dataclass
-from typing import (TYPE_CHECKING, Callable, Dict, Mapping, Optional, Protocol,
-                    Tuple, Type)
+from typing import (TYPE_CHECKING, Any, Callable, Dict, Mapping, Optional,
+                    Protocol, Tuple, Type)
 
 from loguru import logger
 from torch import nn
@@ -49,6 +49,13 @@ def get_hf_config(self, hf_config_type: Type[C] = PretrainedConfig) -> C:
 
         return hf_config
 
+    def get_hf_image_processor_config(self) -> Dict[str, Any]:
+        """
+        Get the HuggingFace image processor configuration of the model.
+        """
+        return self.model_config.hf_image_processor_config
+
+
 
 N = TypeVar("N", bound=Type[nn.Module])
 
diff --git a/aphrodite/modeling/models/phi3v.py b/aphrodite/modeling/models/phi3v.py
index 7d47ae3a3..eeec38c4b 100644
--- a/aphrodite/modeling/models/phi3v.py
+++ b/aphrodite/modeling/models/phi3v.py
@@ -16,8 +16,8 @@
 # limitations under the License.
 import re
 from functools import lru_cache
-from typing import (Iterable, List, Literal, Mapping, Optional, Tuple,
-                    TypedDict, Union)
+from typing import (Any, Dict, Iterable, List, Literal, Mapping, Optional,
+                    Tuple, TypedDict, Union)
 
 import numpy as np
 import torch
@@ -320,12 +320,12 @@ def _calc_hd_transform_size(*, width: int, height: int, hd_num: int = 16):
 
 # Based on https://huggingface.co/microsoft/Phi-3-vision-128k-instruct/blob/main/image_processing_phi3_v.py#L181
 def get_phi3v_image_feature_size(
-    hf_config: PretrainedConfig,
+    hf_config: Dict[str, Any],
     *,
     input_height: int,
     input_width: int,
 ) -> int:
-    num_crops = getattr(hf_config, "num_crops", 16)
+    num_crops = hf_config.get("num_crops", 16)
     new_width, new_height = _calc_hd_transform_size(width=input_width,
                                                     height=input_height,
                                                     hd_num=num_crops)
@@ -337,7 +337,7 @@ def get_phi3v_image_feature_size(
 def get_max_phi3v_image_tokens(ctx: InputContext):
 
     return get_phi3v_image_feature_size(
-        ctx.get_hf_config(PretrainedConfig),
+        ctx.get_hf_image_processor_config(),
         input_height=MAX_IMAGE_FEATURE_SIZE_HEIGHT,
         input_width=MAX_IMAGE_FEATURE_SIZE_WIDTH,
     )
@@ -391,7 +391,7 @@ def input_processor_for_phi3v(ctx: InputContext, llm_inputs: LLMInputs):
         return llm_inputs
 
     model_config = ctx.model_config
-    hf_config = ctx.get_hf_config(PretrainedConfig)
+    hf_config = ctx.get_hf_image_processor_config()
 
     image_data = multi_modal_data["image"]
     if isinstance(image_data, Image.Image):
diff --git a/aphrodite/transformers_utils/config.py b/aphrodite/transformers_utils/config.py
index 3a8d95186..83ff23f09 100644
--- a/aphrodite/transformers_utils/config.py
+++ b/aphrodite/transformers_utils/config.py
@@ -9,6 +9,8 @@
                              try_to_load_from_cache)
 from loguru import logger
 from transformers import GenerationConfig, PretrainedConfig
+from transformers.models.auto.image_processing_auto import (
+    get_image_processor_config)
 from transformers.models.auto.modeling_auto import (
     MODEL_FOR_CAUSAL_LM_MAPPING_NAMES)
 from transformers.utils import CONFIG_NAME as HF_CONFIG_NAME
@@ -243,6 +245,17 @@ def recurse_elems(elem: Any):
     return config
 
 
+def get_hf_image_processor_config(
+    model: Union[str, Path],
+    revision: Optional[str] = None,
+    **kwargs,
+) -> Dict[str, Any]:
+    # Separate model folder from file path for GGUF models
+    if Path(model).is_file() and Path(model).suffix == ".gguf":
+        model = Path(model).parent
+    return get_image_processor_config(model, revision=revision, **kwargs)
+
+
 def get_hf_text_config(config: PretrainedConfig):
     """Get the "sub" config relevant to llm for multi modal models.
         No op for pure text models.
diff --git a/examples/vision/vision_example.py b/examples/vision/vision_example.py
index df6e3c501..51646bdec 100644
--- a/examples/vision/vision_example.py
+++ b/examples/vision/vision_example.py
@@ -58,7 +58,7 @@ def run_phi3v(question):
     # In this example, we override max_num_seqs to 5 while
     # keeping the original context length of 128k.
     llm = LLM(
-        model="microsoft/Phi-3-vision-128k-instruct",
+        model="microsoft/Phi-3.5-vision-instruct",
         trust_remote_code=True,
         max_num_seqs=5,
     )
diff --git a/tests/models/test_phi3v.py b/tests/models/test_phi3v.py
index a55454b8a..9d496eb46 100644
--- a/tests/models/test_phi3v.py
+++ b/tests/models/test_phi3v.py
@@ -21,7 +21,7 @@
     "<|user|>\n<|image_1|>\nWhat is the season?<|end|>\n<|assistant|>\n",
 })
 
-models = ["microsoft/Phi-3-vision-128k-instruct"]
+models = ["microsoft/Phi-3.5-vision-instruct"]
 
 
 def aphrodite_to_hf_output(aphrodite_output: Tuple[List[int], str,