diff --git a/pkgs/development/python-modules/curated-tokenizers/default.nix b/pkgs/development/python-modules/curated-tokenizers/default.nix
new file mode 100644
index 000000000000..77f4a77f906b
--- /dev/null
+++ b/pkgs/development/python-modules/curated-tokenizers/default.nix
@@ -0,0 +1,56 @@
+{
+  lib,
+  buildPythonPackage,
+  fetchFromGitHub,
+  cython,
+  setuptools,
+  regex,
+  pytestCheckHook,
+}:
+
+buildPythonPackage rec {
+  pname = "curated-tokenizers";
+  version = "0.0.9";
+  pyproject = true;
+
+  src = fetchFromGitHub {
+    owner = "explosion";
+    repo = "curated-tokenizers";
+    tag = "v${version}";
+    hash = "sha256-P8kpPnaU3el7sc/vUn4waQN+JV7F9b49i6BtC4BFfIg=";
+    fetchSubmodules = true;
+  };
+
+  build-system = [
+    cython
+    setuptools
+  ];
+
+  dependencies = [
+    regex
+  ];
+
+  nativeCheckInputs = [
+    pytestCheckHook
+  ];
+
+  # Explicitly set the path to avoid running vendored
+  # sentencepiece tests.
+  pytestFlagsArray = [ "tests" ];
+
+  preCheck = ''
+    # avoid local paths, relative imports wont resolve correctly
+    mv curated_tokenizers/tests tests
+    rm -r curated_tokenizers
+  '';
+
+  pythonImportsCheck = [ "curated_tokenizers" ];
+
+  meta = with lib; {
+    description = "Lightweight piece tokenization library";
+    homepage = "https://github.com/explosion/curated-tokenizers";
+    changelog = "https://github.com/explosion/curated-tokenizers/releases/tag/v${version}";
+    license = licenses.mit;
+    maintainers = with maintainers; [ danieldk ];
+  };
+}
diff --git a/pkgs/top-level/python-packages.nix b/pkgs/top-level/python-packages.nix
index bfc7dbd61102..585511d44bf1 100644
--- a/pkgs/top-level/python-packages.nix
+++ b/pkgs/top-level/python-packages.nix
@@ -2791,6 +2791,8 @@ self: super: with self; {
 
   cu2qu = callPackage ../development/python-modules/cu2qu { };
 
+  curated-tokenizers = callPackage ../development/python-modules/curated-tokenizers { };
+
   customtkinter = callPackage ../development/python-modules/customtkinter { };
 
   cucumber-tag-expressions = callPackage ../development/python-modules/cucumber-tag-expressions { };