diff --git a/p256/src/arithmetic/scalar.rs b/p256/src/arithmetic/scalar.rs
index 3a766cb4c..3d741bdeb 100644
--- a/p256/src/arithmetic/scalar.rs
+++ b/p256/src/arithmetic/scalar.rs
@@ -13,7 +13,7 @@ use core::{
 };
 use elliptic_curve::{
     Curve,
-    bigint::{Limb, U256, prelude::*},
+    bigint::{Limb, U256, U384, U512, prelude::*},
     group::ff::{self, Field, PrimeField},
     ops::{Invert, Reduce, ReduceNonZero},
     rand_core::TryRngCore,
@@ -635,6 +635,20 @@ impl Reduce<FieldBytes> for Scalar {
     }
 }
 
+impl Reduce<U512> for Scalar {
+    fn reduce(w: &U512) -> Self {
+        let (lo, hi) = w.split();
+        let w_reduced = barrett_reduce(lo, hi);
+        Self(w_reduced)
+    }
+}
+
+impl Reduce<U384> for Scalar {
+    fn reduce(w: &U384) -> Self {
+        <Self as Reduce<U512>>::reduce(&w.concat(&U128::ZERO))
+    }
+}
+
 impl ReduceNonZero<U256> for Scalar {
     fn reduce_nonzero(w: &U256) -> Self {
         const ORDER_MINUS_ONE: U256 = NistP256::ORDER.wrapping_sub(&U256::ONE);
@@ -651,6 +665,29 @@ impl ReduceNonZero<FieldBytes> for Scalar {
     }
 }
 
+impl ReduceNonZero<U512> for Scalar {
+    fn reduce_nonzero(w: &U512) -> Self {
+        // Reduce U512 to U256 first, then apply non-zero reduction
+        let reduced = <Self as Reduce<U512>>::reduce(w);
+        const ORDER_MINUS_ONE: U256 = NistP256::ORDER.wrapping_sub(&U256::ONE);
+        let (r, underflow) = reduced.0.borrowing_sub(&ORDER_MINUS_ONE, Limb::ZERO);
+        let underflow = Choice::from((underflow.0 >> (Limb::BITS - 1)) as u8);
+        Self(U256::conditional_select(&reduced.0, &r, !underflow).wrapping_add(&U256::ONE))
+    }
+}
+
+impl ReduceNonZero<U384> for Scalar {
+    fn reduce_nonzero(w: &U384) -> Self {
+        // Convert U384 to U512 by zero-padding the high bits
+        let w_bytes = w.to_be_bytes();
+        let mut w512_bytes = [0u8; 64];
+        // Copy U384 bytes to the lower part of U512 (384 bits = 48 bytes)
+        w512_bytes[16..64].copy_from_slice(&w_bytes);
+        let w512 = U512::from_be_byte_array(w512_bytes.into());
+        <Self as ReduceNonZero<U512>>::reduce_nonzero(&w512)
+    }
+}
+
 impl Sum for Scalar {
     fn sum<I: Iterator<Item = Self>>(iter: I) -> Self {
         iter.reduce(Add::add).unwrap_or(Self::ZERO)
@@ -719,7 +756,7 @@ mod tests {
     use crate::{FieldBytes, NistP256, NonZeroScalar, SecretKey};
     use elliptic_curve::{
         Curve,
-        array::Array,
+        bigint::ArrayEncoding,
         group::ff::{Field, PrimeField},
         ops::{BatchInvert, ReduceNonZero},
     };
@@ -785,40 +822,102 @@ mod tests {
 
     #[test]
     fn reduce_nonzero() {
-        assert_eq!(Scalar::reduce_nonzero(&Array::default()).0, U256::ONE,);
-        assert_eq!(Scalar::reduce_nonzero(&U256::ONE).0, U256::from_u8(2),);
         assert_eq!(
-            Scalar::reduce_nonzero(&U256::from_u8(2)).0,
+            <Scalar as ReduceNonZero<U256>>::reduce_nonzero(&U256::ZERO).0,
+            U256::ONE,
+        );
+        assert_eq!(
+            <Scalar as ReduceNonZero<U256>>::reduce_nonzero(&U256::ONE).0,
+            U256::from_u8(2),
+        );
+        assert_eq!(
+            <Scalar as ReduceNonZero<U256>>::reduce_nonzero(&U256::from_u8(2)).0,
             U256::from_u8(3),
         );
 
-        assert_eq!(Scalar::reduce_nonzero(&NistP256::ORDER).0, U256::from_u8(2),);
         assert_eq!(
-            Scalar::reduce_nonzero(&NistP256::ORDER.wrapping_sub(&U256::from_u8(1))).0,
+            <Scalar as ReduceNonZero<U256>>::reduce_nonzero(&NistP256::ORDER).0,
+            U256::from_u8(2),
+        );
+        assert_eq!(
+            <Scalar as ReduceNonZero<U256>>::reduce_nonzero(
+                &NistP256::ORDER.wrapping_sub(&U256::from_u8(1))
+            )
+            .0,
             U256::ONE,
         );
         assert_eq!(
-            Scalar::reduce_nonzero(&NistP256::ORDER.wrapping_sub(&U256::from_u8(2))).0,
+            <Scalar as ReduceNonZero<U256>>::reduce_nonzero(
+                &NistP256::ORDER.wrapping_sub(&U256::from_u8(2))
+            )
+            .0,
             NistP256::ORDER.wrapping_sub(&U256::ONE),
         );
         assert_eq!(
-            Scalar::reduce_nonzero(&NistP256::ORDER.wrapping_sub(&U256::from_u8(3))).0,
+            <Scalar as ReduceNonZero<U256>>::reduce_nonzero(
+                &NistP256::ORDER.wrapping_sub(&U256::from_u8(3))
+            )
+            .0,
             NistP256::ORDER.wrapping_sub(&U256::from_u8(2)),
         );
 
         assert_eq!(
-            Scalar::reduce_nonzero(&NistP256::ORDER.wrapping_add(&U256::ONE)).0,
+            <Scalar as ReduceNonZero<U256>>::reduce_nonzero(
+                &NistP256::ORDER.wrapping_add(&U256::ONE)
+            )
+            .0,
             U256::from_u8(3),
         );
         assert_eq!(
-            Scalar::reduce_nonzero(&NistP256::ORDER.wrapping_add(&U256::from_u8(2))).0,
+            <Scalar as ReduceNonZero<U256>>::reduce_nonzero(
+                &NistP256::ORDER.wrapping_add(&U256::from_u8(2))
+            )
+            .0,
             U256::from_u8(4),
         );
     }
 
+    #[test]
+    fn reduce_nonzero_u384() {
+        use elliptic_curve::bigint::{ArrayEncoding, U384};
+
+        // Test with 48 zero bytes (384 bits)
+        let zero_u384 = U384::ZERO;
+        assert_eq!(
+            <Scalar as ReduceNonZero<U384>>::reduce_nonzero(&zero_u384).0,
+            U256::ONE
+        );
+
+        // Test with small values
+        let mut bytes = [0u8; 48];
+        bytes[47] = 1; // Set the least significant byte to 1
+        let u384_val = U384::from_be_byte_array(bytes.into());
+        assert_eq!(
+            <Scalar as ReduceNonZero<U384>>::reduce_nonzero(&u384_val).0,
+            U256::from_u8(2)
+        );
+
+        bytes[47] = 2;
+        let u384_val2 = U384::from_be_byte_array(bytes.into());
+        assert_eq!(
+            <Scalar as ReduceNonZero<U384>>::reduce_nonzero(&u384_val2).0,
+            U256::from_u8(3)
+        );
+
+        // Test with a value that spans the full 384 bits
+        let large_value = U384::from_be_hex(
+            "FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF",
+        );
+        let reduced = <Scalar as ReduceNonZero<U384>>::reduce_nonzero(&large_value);
+        // The result should be non-zero and within the field
+        assert_ne!(reduced.0, U256::ZERO);
+        assert!(reduced.0 < NistP256::ORDER);
+    }
+
     prop_compose! {
         fn non_zero_scalar()(bytes in any::<[u8; 32]>()) -> NonZeroScalar {
-            NonZeroScalar::reduce_nonzero(&FieldBytes::from(bytes))
+            let uint = U256::from_be_byte_array(bytes.into());
+            <NonZeroScalar as ReduceNonZero<U256>>::reduce_nonzero(&uint)
         }
     }