diff --git a/CHANGELOG.md b/CHANGELOG.md
index 14862c84..e987fea5 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -46,6 +46,7 @@ All notable changes to this project will be documented in this file.
 
 * Reduce intermediate payload copies in bounded MPSC receives, especially for large inline messages.
 * Transfer reclaimed broadcast head messages directly to receivers, avoiding an extra payload `Arc` clone and unnecessary blocked-sender bookkeeping.
+* Reduce the cost of acquiring and releasing an uncontended `Mutex` by avoiding its internal wait-queue lock.
 * Reduce the cost of initializing an uncontended `OnceCell` by avoiding its internal wait-queue lock.
 * Reduce the cost of polling a pending oneshot receive again from the same task, which now keeps its registered waker instead of cloning and dropping it.
 * Reduce the state kept by futures waiting for a `Mutex`, `RwLock`, `Semaphore`, `Once`, `OnceCell`, or bounded broadcast capacity, making lock, permit acquisition, and bounded broadcast send futures smaller.
diff --git a/asyncband/src/internal/semaphore.rs b/asyncband/src/internal/semaphore.rs
index 2964757a..68d5785b 100644
--- a/asyncband/src/internal/semaphore.rs
+++ b/asyncband/src/internal/semaphore.rs
@@ -164,17 +164,6 @@ impl Semaphore {
         self.distribute_permits(n, self.waiters.lock());
     }
 
-    /// Releases `n` permits by acquiring the queue lock directly.
-    ///
-    /// Use this when no permits can be available, as when a mutex guard holds its only permit,
-    /// to skip the speculative attempt in [`release`](Self::release).
-    /// This method also works when permits are available.
-    pub fn release_with_lock(&self, n: usize) {
-        if n != 0 {
-            self.distribute_permits(n, self.waiters.lock());
-        }
-    }
-
     /// Releases `n` permits only if the queue contains waiters when locked.
     #[cfg(feature = "broadcast")]
     pub fn release_if_waiters(&self, n: usize) {
@@ -533,24 +522,6 @@ mod tests {
         assert!(semaphore.waiters.lock().is_empty());
     }
 
-    #[test]
-    fn release_with_lock_hands_permits_to_waiters() {
-        let semaphore = Semaphore::new(1);
-        let counter = Arc::new(WakeCounter(AtomicUsize::new(0)));
-        let waker = Waker::from(counter.clone());
-        assert!(semaphore.try_acquire(1));
-        let mut acquire = semaphore.acquire(1);
-        assert!(acquire.poll_once(&waker).is_pending());
-
-        semaphore.release_with_lock(1);
-        assert_eq!(counter.0.load(Ordering::Relaxed), 1);
-        assert_eq!(semaphore.available_permits(), 0);
-        assert!(acquire.poll_once(&waker).is_ready());
-
-        semaphore.release_with_lock(1);
-        assert_eq!(semaphore.available_permits(), 1);
-    }
-
     #[test]
     fn release_distributes_permits_across_multiple_batches() {
         const WAITER_COUNT: usize = WakerBatch::STACK_SIZE * 2 + 1;
diff --git a/asyncband/src/mutex/mod.rs b/asyncband/src/mutex/mod.rs
index da8e81c2..2b723a2f 100644
--- a/asyncband/src/mutex/mod.rs
+++ b/asyncband/src/mutex/mod.rs
@@ -120,7 +120,9 @@ impl<T> Mutex<T> {
     /// let mutex = Mutex::new(5);
     /// ```
     pub const fn new(t: T) -> Self {
-        let s = semaphore::Semaphore::new(1);
+        // Each lock takes two of three permits, so only one can run at a time. The spare
+        // permit lets an uncontended release avoid locking the semaphore's wait queue.
+        let s = semaphore::Semaphore::new(3);
         let c = UnsafeCell::new(t);
         Self { s, c }
     }
@@ -162,7 +164,7 @@ impl<T: ?Sized> Mutex<T> {
     /// # }
     /// ```
     pub async fn lock(&self) -> MutexGuard<'_, T> {
-        self.s.acquire(1).await;
+        self.s.acquire(2).await;
         MutexGuard { lock: self }
     }
 
@@ -179,7 +181,7 @@ impl<T: ?Sized> Mutex<T> {
     /// assert_eq!(2, *guard);
     /// ```
     pub fn try_lock(&self) -> Option<MutexGuard<'_, T>> {
-        if self.s.try_acquire(1) {
+        if self.s.try_acquire(2) {
             let guard = MutexGuard { lock: self };
             Some(guard)
         } else {
@@ -212,7 +214,7 @@ impl<T: ?Sized> Mutex<T> {
     /// # }
     /// ```
     pub async fn lock_owned(self: Arc<Self>) -> OwnedMutexGuard<T> {
-        self.s.acquire(1).await;
+        self.s.acquire(2).await;
         OwnedMutexGuard { lock: self }
     }
 
@@ -233,7 +235,7 @@ impl<T: ?Sized> Mutex<T> {
     /// assert_eq!(2, *guard);
     /// ```
     pub fn try_lock_owned(self: Arc<Self>) -> Option<OwnedMutexGuard<T>> {
-        if self.s.try_acquire(1) {
+        if self.s.try_acquire(2) {
             let guard = OwnedMutexGuard { lock: self };
             Some(guard)
         } else {
@@ -278,7 +280,7 @@ unsafe impl<T: ?Sized + Send + Sync> Sync for MutexGuard<'_, T> {}
 
 impl<T: ?Sized> Drop for MutexGuard<'_, T> {
     fn drop(&mut self) {
-        self.lock.s.release_with_lock(1);
+        self.lock.s.release(2);
     }
 }
 
@@ -445,7 +447,7 @@ unsafe impl<T: ?Sized + Send + Sync> Sync for OwnedMutexGuard<T> {}
 
 impl<T: ?Sized> Drop for OwnedMutexGuard<T> {
     fn drop(&mut self) {
-        self.lock.s.release_with_lock(1);
+        self.lock.s.release(2);
     }
 }
 
@@ -641,7 +643,7 @@ unsafe impl<T: ?Sized + Sync> Sync for MappedMutexGuard<'_, T> {}
 
 impl<T: ?Sized> Drop for MappedMutexGuard<'_, T> {
     fn drop(&mut self) {
-        self.s.release_with_lock(1);
+        self.s.release(2);
     }
 }
 
@@ -840,7 +842,7 @@ unsafe impl<T: ?Sized + Send + Sync, U: ?Sized + Send + Sync> Sync for OwnedMapp
 
 impl<T: ?Sized, U: ?Sized> Drop for OwnedMappedMutexGuard<T, U> {
     fn drop(&mut self) {
-        self.lock.s.release_with_lock(1);
+        self.lock.s.release(2);
     }
 }