diff --git a/.github/workflows/llm_unit_tests.yml b/.github/workflows/llm_unit_tests.yml
index 69a91566..905ba5a3 100644
--- a/.github/workflows/llm_unit_tests.yml
+++ b/.github/workflows/llm_unit_tests.yml
@@ -76,6 +76,7 @@ jobs:
         shell: bash
         run: |
           echo "SPEECH_DATASET_PATH=${DATASET_DIR}/librispeech_asr_dummy" >> "$GITHUB_ENV"
+          echo "COMMON_VOICE_PATH=${DATASET_DIR}/common_voice" >> "$GITHUB_ENV"
 
           echo "LLAMA_ORIGIN_PATH=${ORIGIN_DIR}/llama-7b-hf" >> "$GITHUB_ENV"
           echo "BLOOM_ORIGIN_PATH=${ORIGIN_DIR}/bloom-7b1" >> "$GITHUB_ENV"
@@ -160,6 +161,11 @@ jobs:
             echo "wget -r -nH --no-verbose --cut-dirs=2 $LLM_FTP_URL/llm/datasets/librispeech_asr_dummy -P $DATASET_DIR"
             wget -r -nH --no-verbose --cut-dirs=2 $LLM_FTP_URL/llm/datasets/librispeech_asr_dummy -P $DATASET_DIR
           fi
+          if [ ! -d $COMMON_VOICE_PATH ]; then
+            echo "Directory $COMMON_VOICE_PATH not found. Downloading from FTP server..."
+            echo "wget -r -nH --no-verbose --cut-dirs=2 $LLM_FTP_URL/llm/datasets/common_voice -P $DATASET_DIR"
+            wget -r -nH --no-verbose --cut-dirs=2 $LLM_FTP_URL/llm/datasets/common_voice -P $DATASET_DIR
+          fi
 
       - name: Run LLM cli test (Linux)
         if: runner.os == 'Linux' 
@@ -167,13 +173,11 @@ jobs:
       - name: Run LLM cli test (Windows)
         if: runner.os == 'Windows' 
         uses: ./.github/actions/llm/cli-test-windows
-        
       - name: Run LLM inference test
         shell: bash
         run: |
-          python -m pip install einops datasets librosa
+          python -m pip install einops datasets librosa openai-whisper
           bash python/llm/test/run-llm-inference-tests.sh
-
       - name: Run LLM langchain test
         shell: bash
         run: |
diff --git a/python/llm/src/bigdl/llm/__init__.py b/python/llm/src/bigdl/llm/__init__.py
index a9c71a16..815be311 100644
--- a/python/llm/src/bigdl/llm/__init__.py
+++ b/python/llm/src/bigdl/llm/__init__.py
@@ -20,3 +20,4 @@
 # only search the first bigdl package and end up finding only one sub-package.
 
 from .convert_model import llm_convert
+from .optimize import optimize_model 
diff --git a/python/llm/src/bigdl/llm/optimize.py b/python/llm/src/bigdl/llm/optimize.py
new file mode 100644
index 00000000..334ad0e9
--- /dev/null
+++ b/python/llm/src/bigdl/llm/optimize.py
@@ -0,0 +1,37 @@
+#
+# Copyright 2016 The BigDL Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+from .transformers import ggml_convert_quant
+from bigdl.llm.ggml.quantize import ggml_tensor_qtype
+from bigdl.llm.utils.common import invalidInputError
+
+
+def optimize_model(model, low_bit='sym_int4', optimize_llm=True):
+    """
+    A method to optimize any pytorch models.
+
+    :param model: The original PyTorch model (nn.module)
+    :param low_bit: Supported low-bit options are "sym_int4", "asym_int4", "sym_int5",
+        "asym_int5" or "sym_int8".
+    :param optimize_llm: Whether to further optimize llm model.
+
+    return: The optimized model.
+    """
+    invalidInputError(low_bit in ggml_tensor_qtype,
+                      f"Unknown load_in_low_bit value: {low_bit}, expected:"
+                      f" sym_int4, asym_int4, sym_int5, asym_int5 or sym_int8.")
+    qtype = ggml_tensor_qtype[low_bit]
+    return ggml_convert_quant(model, qtype=qtype, optimize_model=optimize_llm)
diff --git a/python/llm/test/inference/test_optimize.py b/python/llm/test/inference/test_optimize.py
new file mode 100644
index 00000000..ff1269e5
--- /dev/null
+++ b/python/llm/test/inference/test_optimize.py
@@ -0,0 +1,44 @@
+#
+# Copyright 2016 The BigDL Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+import unittest
+import os
+import pytest
+import time
+import torch
+from bigdl.llm import optimize_model
+        
+class TestOptimizeAPI(unittest.TestCase):
+
+    def setUp(self):   
+        thread_num = os.environ.get('THREAD_NUM')
+        if thread_num is not None:
+            self.n_threads = int(thread_num)
+        else:
+            self.n_threads = 2
+    
+    def test_optimize_whisper(self):
+        # dataset_path = os.environ.get('COMMON_VOICE_PATH')
+        # reservation_audio = os.path.join(dataset_path,'reservation.mp3')
+        import whisper
+        model = whisper.load_model("tiny")
+        model = optimize_model(model, low_bit="sym_int4", optimize_llm=False)
+        # result = model.transcribe(reservation_audio, verbose=True, language="English")
+        # assert "Reservation" or "reservation" in result["text"]
+        
+        
+if __name__ == '__main__':
+    pytest.main([__file__])
\ No newline at end of file