diff --git a/sqlmesh/utils/__init__.py b/sqlmesh/utils/__init__.py
index c220de4847..5b1b077216 100644
--- a/sqlmesh/utils/__init__.py
+++ b/sqlmesh/utils/__init__.py
@@ -21,6 +21,7 @@
 from functools import lru_cache, reduce, wraps
 from pathlib import Path
 
+import unicodedata
 from sqlglot import exp
 from sqlglot.dialects.dialect import Dialects
 
@@ -291,8 +292,14 @@ def sqlglot_dialects() -> str:
 
 NON_ALNUM = re.compile(r"[^a-zA-Z0-9_]")
 
+NON_ALUM_INCLUDE_UNICODE = re.compile(r"\W", flags=re.UNICODE)
 
-def sanitize_name(name: str) -> str:
+
+def sanitize_name(name: str, *, include_unicode: bool = False) -> str:
+    if include_unicode:
+        s = unicodedata.normalize("NFC", name)
+        s = NON_ALUM_INCLUDE_UNICODE.sub("_", s)
+        return s
     return NON_ALNUM.sub("_", name)
 
 
diff --git a/sqlmesh/utils/cache.py b/sqlmesh/utils/cache.py
index 002248f511..4b557e43b6 100644
--- a/sqlmesh/utils/cache.py
+++ b/sqlmesh/utils/cache.py
@@ -133,7 +133,7 @@ def clear(self) -> None:
 
     def _cache_entry_path(self, name: str, entry_id: str = "") -> Path:
         entry_file_name = "__".join(p for p in (self._cache_version, name, entry_id) if p)
-        full_path = self._path / sanitize_name(entry_file_name)
+        full_path = self._path / sanitize_name(entry_file_name, include_unicode=True)
         if IS_WINDOWS:
             # handle paths longer than 260 chars
             full_path = fix_windows_path(full_path)
diff --git a/tests/utils/__init__.py b/tests/utils/__init__.py
index e69de29bb2..744ad37757 100644
--- a/tests/utils/__init__.py
+++ b/tests/utils/__init__.py
@@ -0,0 +1,23 @@
+import pytest
+
+from sqlmesh.utils import sanitize_name
+
+
+@pytest.mark.parametrize(
+    "raw,exclude_unicode,include_unicode",
+    [
+        ("simple", "simple", "simple"),
+        ("snake_case", "snake_case", "snake_case"),
+        ("客户数据", "____", "客户数据"),
+        ("客户-数据 v2", "______v2", "客户_数据_v2"),
+        ("中文，逗号", "_____", "中文_逗号"),
+        ("a/b", "a_b", "a_b"),
+        ("spaces\tand\nnewlines", "spaces_and_newlines", "spaces_and_newlines"),
+        ("data📦2025", "data_2025", "data_2025"),
+        ("MiXeD123_名字", "MiXeD123___", "MiXeD123_名字"),
+        ("", "", ""),
+    ],
+)
+def test_sanitize_name_no_(raw, exclude_unicode, include_unicode):
+    assert sanitize_name(raw) == exclude_unicode
+    assert sanitize_name(raw, include_unicode=True) == include_unicode
diff --git a/tests/utils/test_cache.py b/tests/utils/test_cache.py
index cd1fdb0115..0b6d335446 100644
--- a/tests/utils/test_cache.py
+++ b/tests/utils/test_cache.py
@@ -39,6 +39,7 @@ def test_file_cache(tmp_path: Path, mocker: MockerFixture):
     loader.assert_called_once()
 
     assert "___test_model_" in cache._cache_entry_path('"test_model"').name
+    assert "客户数据" in cache._cache_entry_path("客户数据").name
 
 
 def test_optimized_query_cache(tmp_path: Path, mocker: MockerFixture):