diff --git a/plonky2/Cargo.toml b/plonky2/Cargo.toml
index 94715189b3..791a64c922 100644
--- a/plonky2/Cargo.toml
+++ b/plonky2/Cargo.toml
@@ -80,6 +80,10 @@ harness = false
 name = "ffts"
 harness = false
 
+[[bench]]
+name = "lde"
+harness = false
+
 [[bench]]
 name = "hashing"
 harness = false
diff --git a/plonky2/benches/lde.rs b/plonky2/benches/lde.rs
new file mode 100644
index 0000000000..c1a702f931
--- /dev/null
+++ b/plonky2/benches/lde.rs
@@ -0,0 +1,46 @@
+mod allocator;
+
+use criterion::{criterion_group, criterion_main, BenchmarkId, Criterion};
+use plonky2::field::extension::Extendable;
+use plonky2::field::goldilocks_field::GoldilocksField;
+use plonky2::field::polynomial::PolynomialCoeffs;
+use plonky2::fri::oracle::PolynomialBatch;
+use plonky2::hash::hash_types::RichField;
+use plonky2::plonk::config::{GenericConfig, PoseidonGoldilocksConfig};
+use plonky2::util::timing::TimingTree;
+use tynm::type_name;
+#[cfg(feature = "cuda")]
+use cryptography_cuda::{init_cuda_degree_rs};
+
+pub(crate) fn bench_batch_lde<F: RichField + Extendable<D>, C: GenericConfig<D, F = F>, const D: usize>(c: &mut Criterion)
+{
+    const RATE_BITS: usize = 3;
+
+    let mut group = c.benchmark_group(&format!("lde<{}>", type_name::<F>()));
+
+    #[cfg(feature = "cuda")]
+    init_cuda_degree_rs(16);
+
+    for size_log in [13, 14, 15] {
+        let orig_size = 1 << (size_log - RATE_BITS);
+        let lde_size = 1 << size_log;
+        let batch_size = 1 << 4;
+
+        group.bench_with_input(BenchmarkId::from_parameter(lde_size), &lde_size, |b, _| {
+            let polynomials: Vec<PolynomialCoeffs<F>> = (0..batch_size).into_iter().map(|_i| {
+                PolynomialCoeffs::new(F::rand_vec(orig_size))
+            }).collect();
+            let mut timing = TimingTree::new("lde", log::Level::Error);
+            b.iter(|| {
+                PolynomialBatch::<F, C, D>::from_coeffs(polynomials.clone(), RATE_BITS, false, 1, &mut timing, None)
+            });
+        });
+    }
+}
+
+fn criterion_benchmark(c: &mut Criterion) {
+    bench_batch_lde::<GoldilocksField, PoseidonGoldilocksConfig, 2>(c);
+}
+
+criterion_group!(benches, criterion_benchmark);
+criterion_main!(benches);