diff --git a/Cargo.toml b/Cargo.toml
index 86c98cf1e..1c84ad8db 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -50,6 +50,11 @@ name = "deku"
 harness = false
 required-features = ["alloc"]
 
+[[bench]]
+name = "bebits"
+harness = false
+required-features = ["alloc", "bits"]
+
 [lints]
 workspace = true
 
diff --git a/benches/bebits.rs b/benches/bebits.rs
new file mode 100644
index 000000000..86ed0bcbe
--- /dev/null
+++ b/benches/bebits.rs
@@ -0,0 +1,75 @@
+//! Big-endian bit-packed headers: the shape used by real network / space
+//! protocols (CCSDS, IPv4, DVB-S2). Mirrors the CCSDS TM Transfer Frame
+//! primary header, 6 octets / 11 fields.
+use criterion::{criterion_group, criterion_main, Criterion};
+use deku::prelude::*;
+use no_std_io::io::Cursor;
+
+#[derive(Debug, PartialEq, DekuRead, DekuWrite)]
+#[deku(endian = "big")]
+struct TmPrimaryHeader {
+    #[deku(bits = 2)]
+    tfvn: u8,
+    #[deku(bits = 10)]
+    scid: u16,
+    #[deku(bits = 3)]
+    vcid: u8,
+    #[deku(bits = 1)]
+    ocf: u8,
+    mcfc: u8,
+    vcfc: u8,
+    #[deku(bits = 1)]
+    tfs: u8,
+    #[deku(bits = 1)]
+    syn: u8,
+    #[deku(bits = 1)]
+    po: u8,
+    #[deku(bits = 2)]
+    sli: u8,
+    #[deku(bits = 11)]
+    fhp: u16,
+}
+
+/// Same 6 octets, byte-aligned: the deku fast path, for scale.
+#[derive(Debug, PartialEq, DekuRead, DekuWrite)]
+#[deku(endian = "big")]
+struct SixBytes {
+    a: u8,
+    b: u8,
+    c: u8,
+    d: u8,
+    e: u8,
+    f: u8,
+}
+
+/// Worst case in the docs: a 1-bit field in a wide container.
+#[derive(Debug, PartialEq, DekuRead, DekuWrite)]
+#[deku(endian = "big")]
+struct OneBitU64 {
+    #[deku(bits = 1)]
+    a: u64,
+}
+
+fn bench(c: &mut Criterion) {
+    let buf = [0x2Au8, 0xB5, 0x11, 0x22, 0xC7, 0xFF, 0x00, 0x99];
+    c.bench_function("be_tm_primary_header_11_fields", |b| {
+        b.iter(|| {
+            let mut r = Reader::new(Cursor::new(&buf));
+            TmPrimaryHeader::from_reader_with_ctx(&mut r, ()).unwrap()
+        })
+    });
+    c.bench_function("be_six_bytes_aligned", |b| {
+        b.iter(|| {
+            let mut r = Reader::new(Cursor::new(&buf));
+            SixBytes::from_reader_with_ctx(&mut r, ()).unwrap()
+        })
+    });
+    c.bench_function("be_one_bit_in_u64", |b| {
+        b.iter(|| {
+            let mut r = Reader::new(Cursor::new(&buf));
+            OneBitU64::from_reader_with_ctx(&mut r, ()).unwrap()
+        })
+    });
+}
+criterion_group!(bebits, bench);
+criterion_main!(bebits);
diff --git a/src/impls/primitive.rs b/src/impls/primitive.rs
index d35f8a2a0..edfee1dce 100644
--- a/src/impls/primitive.rs
+++ b/src/impls/primitive.rs
@@ -181,6 +181,22 @@ macro_rules! ImplDekuReadBits {
                     }
                 }
 
+                // Fast path: big-endian, Msb0 bit order.
+                //
+                // The generic paths below zero-extend the value to the container
+                // width by calling `BoundedBitVec::insert(0, false)` once per
+                // padding bit, and each `insert` shifts the entire backing array
+                // right by one bit. That is O(MAX_TYPE_BITS) bit operations per
+                // padding bit, i.e. O(MAX_TYPE_BITS * (MAX_TYPE_BITS - bit_size))
+                // per field -- for a 1-bit field in a u64 that is 63 whole-array
+                // shifts. `load_be` performs the identical zero-extension in one
+                // step. Only Msb0 + big-endian is handled here; the Lsb0 and
+                // little-endian paths keep their existing behaviour.
+                if !input_is_le && order == Order::Msb0 && bit_size > 0 {
+                    let value = bit_slice.load_be::<$inner>();
+                    return Ok((bit_size, <$typ>::from_be_bytes(value.to_be_bytes())));
+                }
+
                 // if read from Lsb order and it's especially cursed since its not just within one byte...
                 // read_bits returned: [0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 1]
                 //                     | second     |    first            |
@@ -286,6 +302,15 @@ macro_rules! ImplDekuReadBits {
                     }
                 }
 
+                // Fast path: big-endian. This impl has no `Order` parameter and
+                // so is always Msb0. See the comment on the same fast path in
+                // the `(Endian, BitSize, Order)` impl above for why the generic
+                // path below is quadratic in the container width.
+                if !input_is_le && bit_size > 0 {
+                    let value = bit_slice.load_be::<$inner>();
+                    return Ok((bit_size, <$typ>::from_be_bytes(value.to_be_bytes())));
+                }
+
                 // Create a new BoundedBitVec from the slice and pad un-aligned chunks
                 // i.e. [10010110, 1110] -> [10010110, 00001110]
                 const MAX_TYPE_BYTES: usize = core::mem::size_of::<$typ>();