diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index 85bfddda..b212ac55 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -36,7 +36,14 @@ jobs:
         if: matrix.platform == 'ubuntu-22.04'
         run: |
           sudo apt-get update
-          sudo apt-get install -y libwebkit2gtk-4.1-dev libappindicator3-dev librsvg2-dev patchelf
+          sudo apt-get install -y libwebkit2gtk-4.1-dev libappindicator3-dev librsvg2-dev patchelf llvm-dev
+
+      - name: Install LLVM (macOS)
+        if: matrix.platform == 'macos-latest' || matrix.platform == 'macos-15-intel'
+        run: |
+          brew install llvm
+          echo "$(brew --prefix llvm)/bin" >> $GITHUB_PATH
+          echo "LLVM_CONFIG=$(brew --prefix llvm)/bin/llvm-config" >> $GITHUB_ENV
 
       - name: Setup Python
         uses: actions/setup-python@v5
diff --git a/backend/build_binary.py b/backend/build_binary.py
index 1b665de6..6fa512eb 100644
--- a/backend/build_binary.py
+++ b/backend/build_binary.py
@@ -26,6 +26,18 @@ def build_server():
         args.extend(['--paths', str(local_qwen_path)])
         print(f"Using local qwen_tts source from: {local_qwen_path}")
 
+    # Exclude unnecessary modules to reduce size
+    args.extend([
+        '--exclude-module', 'matplotlib',
+        '--exclude-module', 'IPython',
+        '--exclude-module', 'notebook',
+        '--exclude-module', 'pytest',
+        '--exclude-module', 'setuptools',
+        '--exclude-module', 'torch.distributions',
+        '--exclude-module', 'torch.testing',
+        '--exclude-module', 'tensorboard',
+    ])
+
     # Add hidden imports
     args.extend([
         '--hidden-import', 'backend',