diff --git a/.github/workflows/test-release.yaml b/.github/workflows/test-release.yaml
index 62753e1d..c0817162 100644
--- a/.github/workflows/test-release.yaml
+++ b/.github/workflows/test-release.yaml
@@ -260,7 +260,7 @@ jobs:
       - name: Setup node
         uses: actions/setup-node@v4
         with:
-          node-version: 20
+          node-version: 22
           cache: 'yarn'
 
       - name: Install
@@ -475,7 +475,7 @@ jobs:
         run: |
           node -e "console.info('docker-image=${{ matrix.settings.image }}'.replace('{:version}', ${{ matrix.node }}))" >> "$GITHUB_OUTPUT"
       - name: Run test
-        run: docker run --rm ${{ matrix.settings.args }} -v ${{ github.workspace }}:/build -w /build ${{ steps.image-name.outputs.docker-image }} sh -c "DISABLE_V8_COMPILE_CACHE=1 yarn workspace @examples/napi test -s"
+        run: docker run --rm ${{ matrix.settings.args }} -v ${{ github.workspace }}:/build -w /build ${{ steps.image-name.outputs.docker-image }} sh -c "NODE_OPTIONS=--max-old-space-size=6000 DISABLE_V8_COMPILE_CACHE=1 yarn workspace @examples/napi test -s"
 
   build-and-test-linux-armv7:
     name: stable - armv7-unknown-linux-gnueabihf - node@20
diff --git a/bench/Cargo.toml b/bench/Cargo.toml
index 8c25a97b..e9f825a1 100644
--- a/bench/Cargo.toml
+++ b/bench/Cargo.toml
@@ -9,7 +9,6 @@ version = "0.1.0"
 crate-type = ["cdylib"]
 
 [dependencies]
-ctor = "0.3"
 napi = { path = "../crates/napi", features = [
   "tokio_rt",
   "serde-json",
diff --git a/crates/backend/src/codegen/const.rs b/crates/backend/src/codegen/const.rs
index 482eb2fc..6a1980c1 100644
--- a/crates/backend/src/codegen/const.rs
+++ b/crates/backend/src/codegen/const.rs
@@ -36,7 +36,7 @@ impl NapiConst {
       #[allow(non_snake_case)]
       #[allow(clippy::all)]
       #[cfg(all(not(test), not(target_family = "wasm")))]
-      #[napi::bindgen_prelude::ctor]
+      #[napi::ctor::ctor(crate_path=::napi::ctor)]
       fn #register_name() {
         napi::bindgen_prelude::register_module_export(#js_mod_ident, #js_name_lit, #cb_name);
       }
diff --git a/crates/backend/src/codegen/enum.rs b/crates/backend/src/codegen/enum.rs
index 69b1acea..d971e658 100644
--- a/crates/backend/src/codegen/enum.rs
+++ b/crates/backend/src/codegen/enum.rs
@@ -254,7 +254,7 @@ impl NapiEnum {
       #[allow(non_snake_case)]
       #[allow(clippy::all)]
       #[cfg(all(not(test), not(target_family = "wasm")))]
-      #[napi::bindgen_prelude::ctor]
+      #[napi::ctor::ctor(crate_path=napi::ctor)]
       fn #register_name() {
         napi::bindgen_prelude::register_module_export(#js_mod_ident, #js_name_lit, #callback_name);
       }
diff --git a/crates/backend/src/codegen/fn.rs b/crates/backend/src/codegen/fn.rs
index 2474096f..0d307843 100644
--- a/crates/backend/src/codegen/fn.rs
+++ b/crates/backend/src/codegen/fn.rs
@@ -702,7 +702,7 @@ impl NapiFn {
         #[allow(clippy::all)]
         #[allow(non_snake_case)]
         #[cfg(all(not(test), not(target_family = "wasm")))]
-        #[napi::bindgen_prelude::ctor]
+        #[napi::ctor::ctor(crate_path=::napi::ctor)]
         fn #module_register_name() {
           napi::bindgen_prelude::register_module_export(#js_mod_ident, #js_name, #cb_name);
         }
diff --git a/crates/backend/src/codegen/struct.rs b/crates/backend/src/codegen/struct.rs
index 4c663641..b95ec93e 100644
--- a/crates/backend/src/codegen/struct.rs
+++ b/crates/backend/src/codegen/struct.rs
@@ -869,7 +869,7 @@ impl NapiStruct {
       #[allow(non_snake_case)]
       #[allow(clippy::all)]
       #[cfg(all(not(test), not(target_family = "wasm")))]
-      #[napi::bindgen_prelude::ctor]
+      #[napi::ctor::ctor(crate_path=napi::ctor)]
       fn #struct_register_name() {
         napi::__private::register_class(std::any::TypeId::of::<#name>(), #js_mod_ident, #js_name, vec![#(#props),*]);
       }
@@ -1402,7 +1402,7 @@ impl NapiImpl {
         #(#methods)*
 
         #[cfg(all(not(test), not(target_family = "wasm")))]
-        #[napi::bindgen_prelude::ctor]
+        #[napi::ctor::ctor(crate_path=napi::ctor)]
         fn #register_name() {
           napi::__private::register_class(std::any::TypeId::of::<#name>(), #js_mod_ident, #js_name, vec![#(#props),*]);
         }
diff --git a/crates/macro/src/lib.rs b/crates/macro/src/lib.rs
index ef34952b..17a2a06e 100644
--- a/crates/macro/src/lib.rs
+++ b/crates/macro/src/lib.rs
@@ -151,7 +151,7 @@ pub fn module_exports(_attr: TokenStream, input: TokenStream) -> TokenStream {
   };
 
   let register = quote! {
-    #[cfg_attr(not(target_family = "wasm"), napi::bindgen_prelude::ctor)]
+    #[cfg_attr(not(target_family = "wasm"), napi::ctor::ctor(crate_path=napi::ctor))]
     fn __napi_explicit_module_register() {
       unsafe fn register(raw_env: napi::sys::napi_env, raw_exports: napi::sys::napi_value) -> napi::Result<()> {
         use napi::{Env, JsObject, NapiValue};
@@ -174,3 +174,13 @@ pub fn module_exports(_attr: TokenStream, input: TokenStream) -> TokenStream {
   })
   .into()
 }
+
+#[proc_macro_attribute]
+pub fn module_init(_: TokenStream, input: TokenStream) -> TokenStream {
+  let input = parse_macro_input!(input as ItemFn);
+  quote! {
+    #[napi::ctor::ctor(crate_path=napi::ctor)]
+    #input
+  }
+  .into()
+}
diff --git a/crates/napi/Cargo.toml b/crates/napi/Cargo.toml
index 1b59688a..0b88e91f 100644
--- a/crates/napi/Cargo.toml
+++ b/crates/napi/Cargo.toml
@@ -62,7 +62,7 @@ node_version_detect = []
 
 [dependencies]
 bitflags = "2"
-ctor = "0.3"
+ctor = "0.3.6"
 
 [dependencies.anyhow]
 optional = true
diff --git a/crates/napi/src/bindgen_runtime/module_register.rs b/crates/napi/src/bindgen_runtime/module_register.rs
index 136663e5..353b202d 100644
--- a/crates/napi/src/bindgen_runtime/module_register.rs
+++ b/crates/napi/src/bindgen_runtime/module_register.rs
@@ -7,7 +7,7 @@ use std::mem::MaybeUninit;
 #[cfg(not(feature = "noop"))]
 use std::ptr;
 #[cfg(not(feature = "noop"))]
-use std::sync::atomic::{AtomicBool, Ordering};
+use std::sync::atomic::{AtomicBool, AtomicUsize, Ordering};
 use std::sync::{LazyLock, RwLock};
 use std::thread::ThreadId;
 use std::{any::TypeId, collections::HashMap};
@@ -52,6 +52,7 @@ impl<K, V> Default for PersistedPerInstanceHashMap<K, V> {
   }
 }
 
+#[cfg(not(feature = "noop"))]
 type ModuleRegisterCallback =
   RwLock<Vec<(Option<&'static str>, (&'static str, ExportRegisterCallback))>>;
 
@@ -65,10 +66,12 @@ type FnRegisterMap =
   PersistedPerInstanceHashMap<ExportRegisterCallback, (sys::napi_callback, &'static str)>;
 type RegisteredClassesMap = PersistedPerInstanceHashMap<ThreadId, RegisteredClasses>;
 
+#[cfg(not(feature = "noop"))]
 static MODULE_REGISTER_CALLBACK: LazyLock<ModuleRegisterCallback> = LazyLock::new(Default::default);
+#[cfg(not(feature = "noop"))]
 static MODULE_CLASS_PROPERTIES: LazyLock<ModuleClassProperty> = LazyLock::new(Default::default);
 #[cfg(not(feature = "noop"))]
-static IS_FIRST_MODULE: AtomicBool = AtomicBool::new(true);
+static MODULE_COUNT: AtomicUsize = AtomicUsize::new(0);
 #[cfg(not(feature = "noop"))]
 static FIRST_MODULE_REGISTERED: AtomicBool = AtomicBool::new(false);
 static REGISTERED_CLASSES: LazyLock<RegisteredClassesMap> = LazyLock::new(Default::default);
@@ -109,6 +112,7 @@ pub fn register_module_exports(callback: ModuleExportsCallback) {
     .push(callback);
 }
 
+#[cfg(not(feature = "noop"))]
 #[doc(hidden)]
 pub fn register_module_export(
   js_mod: Option<&'static str>,
@@ -142,6 +146,7 @@ pub fn get_class_constructor(js_name: &'static str) -> Option<sys::napi_ref> {
   })?
 }
 
+#[cfg(not(feature = "noop"))]
 #[doc(hidden)]
 pub fn register_class(
   rust_type_id: TypeId,
@@ -212,10 +217,7 @@ pub unsafe extern "C" fn napi_register_module_v1(
   env: sys::napi_env,
   exports: sys::napi_value,
 ) -> sys::napi_value {
-  #[cfg(all(
-    any(target_env = "msvc", feature = "dyn-symbols"),
-    not(feature = "noop")
-  ))]
+  #[cfg(any(target_env = "msvc", feature = "dyn-symbols"))]
   unsafe {
     sys::setup();
   }
@@ -234,11 +236,11 @@ pub unsafe extern "C" fn napi_register_module_v1(
       NODE_VERSION_PATCH = node_version.patch;
     };
   }
-  if IS_FIRST_MODULE.load(Ordering::SeqCst) {
-    IS_FIRST_MODULE.store(false, Ordering::SeqCst);
-  } else {
+
+  if MODULE_COUNT.fetch_add(1, Ordering::SeqCst) != 0 {
     wait_first_thread_registered();
   }
+
   let mut exports_objects: HashSet<String> = HashSet::default();
 
   {
@@ -424,29 +426,21 @@ pub unsafe extern "C" fn napi_register_module_v1(
     })
   }
 
+  #[cfg(feature = "napi4")]
   let current_thread_id = std::thread::current().id();
 
-  #[cfg(all(not(target_family = "wasm"), feature = "napi3"))]
+  // attach cleanup hook to the `module` object
+  // we don't use the `napi_add_env_cleanup_hook` because it's required napi3
   check_status_or_throw!(
     env,
     unsafe {
-      sys::napi_add_env_cleanup_hook(
+      sys::napi_wrap(
         env,
-        Some(thread_cleanup),
+        exports,
         Box::into_raw(Box::new(current_thread_id)).cast(),
-      )
-    },
-    "Failed to add remove thread id cleanup hook"
-  );
-
-  #[cfg(all(target_family = "wasm", feature = "napi3"))]
-  check_status_or_throw!(
-    env,
-    unsafe {
-      crate::napi_add_env_cleanup_hook(
-        env,
         Some(thread_cleanup),
-        Box::into_raw(Box::new(current_thread_id)).cast(),
+        ptr::null_mut(),
+        ptr::null_mut(),
       )
     },
     "Failed to add remove thread id cleanup hook"
@@ -538,19 +532,24 @@ fn create_custom_gc(env: sys::napi_env, current_thread_id: ThreadId) {
   THREADS_CAN_ACCESS_ENV.borrow_mut(|m| m.insert(current_thread_id, true));
 }
 
-#[cfg(all(feature = "napi3", not(feature = "noop")))]
-unsafe extern "C" fn thread_cleanup(id: *mut std::ffi::c_void) {
-  let thread_id = unsafe { Box::from_raw(id.cast::<ThreadId>()) };
-  REGISTERED_CLASSES.borrow_mut(|m| {
-    m.remove(&thread_id);
-  });
-  #[cfg(feature = "napi4")]
-  {
-    THREADS_CAN_ACCESS_ENV.borrow_mut(|m| m.remove(&thread_id));
-    #[cfg(feature = "tokio_rt")]
+#[cfg(not(feature = "noop"))]
+unsafe extern "C" fn thread_cleanup(
+  _: sys::napi_env,
+  #[allow(unused_variables)] id: *mut std::ffi::c_void,
+  _data: *mut std::ffi::c_void,
+) {
+  if MODULE_COUNT.fetch_sub(1, Ordering::Relaxed) == 1 {
+    #[cfg(all(feature = "tokio_rt", feature = "napi4"))]
     {
       crate::tokio_runtime::drop_runtime();
     }
+    crate::bindgen_runtime::REFERENCE_MAP.borrow_mut(|m| m.clear());
+    return;
+  }
+  #[cfg(feature = "napi4")]
+  {
+    let thread_id = unsafe { Box::from_raw(id.cast::<ThreadId>()) };
+    THREADS_CAN_ACCESS_ENV.borrow_mut(|m| m.remove(&thread_id));
   }
 }
 
diff --git a/crates/napi/src/lib.rs b/crates/napi/src/lib.rs
index dab1041c..23f1f3a2 100644
--- a/crates/napi/src/lib.rs
+++ b/crates/napi/src/lib.rs
@@ -7,7 +7,7 @@
 //!
 //! ## Feature flags
 //!
-//! ### napi1 ~ napi8
+//! ### napi1 ~ napi9
 //!
 //! Because `Node.js` N-API has versions. So there are feature flags to choose what version of `N-API` you want to build for.
 //! For example, if you want build a library which can be used by `node@10.17.0`, you should choose the `napi5` or lower.
@@ -149,8 +149,6 @@ macro_rules! assert_type_of {
   };
 }
 
-pub use crate::bindgen_runtime::ctor as module_init;
-
 pub mod bindgen_prelude {
   #[cfg(all(feature = "compat-mode", not(feature = "noop")))]
   pub use crate::bindgen_runtime::register_module_exports;
@@ -213,6 +211,8 @@ pub mod __private {
   }
 }
 
+pub extern crate ctor;
+
 #[cfg(feature = "tokio_rt")]
 pub extern crate tokio;
 
diff --git a/crates/napi/src/tokio_runtime.rs b/crates/napi/src/tokio_runtime.rs
index 8ad00fe1..f60c2fff 100644
--- a/crates/napi/src/tokio_runtime.rs
+++ b/crates/napi/src/tokio_runtime.rs
@@ -62,33 +62,22 @@ pub fn create_custom_tokio_runtime(rt: Runtime) {
   USER_DEFINED_RT.get_or_init(move || Mutex::new(Some(rt)));
 }
 
-#[cfg(not(feature = "noop"))]
-static RT_REFERENCE_COUNT: std::sync::atomic::AtomicUsize = std::sync::atomic::AtomicUsize::new(0);
-
 #[cfg(not(feature = "noop"))]
 /// Ensure that the Tokio runtime is initialized.
 /// In windows the Tokio runtime will be dropped when Node env exits.
 /// But in Electron renderer process, the Node env will exits and recreate when the window reloads.
 /// So we need to ensure that the Tokio runtime is initialized when the Node env is created.
 pub(crate) fn ensure_runtime() {
-  use std::sync::atomic::Ordering;
-
   let mut rt = RT.write().unwrap();
   if rt.is_none() {
     *rt = create_runtime();
   }
-
-  RT_REFERENCE_COUNT.fetch_add(1, Ordering::Relaxed);
 }
 
 #[cfg(not(feature = "noop"))]
 pub(crate) fn drop_runtime() {
-  use std::sync::atomic::Ordering;
-
-  if RT_REFERENCE_COUNT.fetch_sub(1, Ordering::AcqRel) == 1 {
-    if let Some(rt) = RT.write().unwrap().take() {
-      rt.shutdown_background();
-    }
+  if let Some(rt) = RT.write().unwrap().take() {
+    rt.shutdown_background();
   }
 }
 
diff --git a/examples/napi-compat-mode/Cargo.toml b/examples/napi-compat-mode/Cargo.toml
index b656d9ae..9f756def 100644
--- a/examples/napi-compat-mode/Cargo.toml
+++ b/examples/napi-compat-mode/Cargo.toml
@@ -14,7 +14,6 @@ napi3 = ["napi/napi3"]
 dyn-symbols = ["napi/dyn-symbols"]
 
 [dependencies]
-ctor = "0.3"
 futures = "0.3"
 napi = { path = "../../crates/napi", features = [
   "tokio_rt",
diff --git a/examples/napi/Cargo.toml b/examples/napi/Cargo.toml
index 1b64b999..664331fe 100644
--- a/examples/napi/Cargo.toml
+++ b/examples/napi/Cargo.toml
@@ -15,7 +15,6 @@ error_try_builds = []
 
 [dependencies]
 chrono = "0.4"
-ctor = "0.3"
 futures = "0.3"
 bytes = "1"
 napi-derive = { path = "../../crates/macro", features = ["type-def"] }
diff --git a/examples/napi/src/lib.rs b/examples/napi/src/lib.rs
index a89b2ce2..da1cfba9 100644
--- a/examples/napi/src/lib.rs
+++ b/examples/napi/src/lib.rs
@@ -18,7 +18,7 @@ extern crate serde_derive;
 static ALLOC: snmalloc_rs::SnMalloc = snmalloc_rs::SnMalloc;
 
 #[cfg(not(target_family = "wasm"))]
-#[napi::module_init]
+#[napi_derive::module_init]
 fn init() {
   let rt = tokio::runtime::Builder::new_multi_thread()
     .enable_all()
diff --git a/memory-testing/Cargo.toml b/memory-testing/Cargo.toml
index 7e3a9f60..26839a62 100644
--- a/memory-testing/Cargo.toml
+++ b/memory-testing/Cargo.toml
@@ -9,7 +9,6 @@ version = "0.1.0"
 crate-type = ["cdylib"]
 
 [dependencies]
-ctor = "0.3"
 futures = "0.3"
 napi = { path = "../crates/napi", features = [
   "tokio_rt",