diff --git a/av1/av1.cmake b/av1/av1.cmake
index 5e730f9..4b92eb4 100644
--- a/av1/av1.cmake
+++ b/av1/av1.cmake
@@ -363,6 +363,7 @@
             "${AOM_ROOT}/av1/encoder/x86/av1_fwd_txfm2d_sse4.c"
             "${AOM_ROOT}/av1/encoder/x86/encodetxb_sse4.c"
             "${AOM_ROOT}/av1/encoder/x86/highbd_fwd_txfm_sse4.c"
+            "${AOM_ROOT}/av1/encoder/x86/model_rd_sse4.c"
             "${AOM_ROOT}/av1/encoder/x86/rdopt_sse4.c"
             "${AOM_ROOT}/av1/encoder/x86/pickrst_sse4.c")
 
diff --git a/av1/common/av1_rtcd_defs.pl b/av1/common/av1_rtcd_defs.pl
index d5d6370..bd59340 100644
--- a/av1/common/av1_rtcd_defs.pl
+++ b/av1/common/av1_rtcd_defs.pl
@@ -460,6 +460,9 @@
   add_proto qw/void av1_get_horver_correlation_full/, "const int16_t *diff, int stride, int w, int h, float *hcorr, float *vcorr";
   specialize qw/av1_get_horver_correlation_full sse4_1 avx2 neon/;
 
+  add_proto qw/void av1_interp_cubic_rate_dist/, "const double *p1, const double *p2, double x, double * const rate_f, double * const distbysse_f";
+  specialize qw/av1_interp_cubic_rate_dist sse4_1/;
+
   add_proto qw/void av1_nn_predict/, "const float *input_nodes, const NN_CONFIG *const nn_config, int reduce_prec, float *const output";
 
   add_proto qw/void av1_nn_fast_softmax_16/, "const float *input_nodes, float *output";
diff --git a/av1/encoder/rd.c b/av1/encoder/rd.c
index e9f6e54..b5f6b3a 100644
--- a/av1/encoder/rd.c
+++ b/av1/encoder/rd.c
@@ -949,6 +949,13 @@
                           x * (3.0 * (p[1] - p[2]) + p[3] - p[0])));
 }
 
+void av1_interp_cubic_rate_dist_c(const double *p1, const double *p2, double x,
+                                  double *const rate_f,
+                                  double *const distbysse_f) {
+  *rate_f = interp_cubic(p1, x);
+  *distbysse_f = interp_cubic(p2, x);
+}
+
 /*
 static double interp_bicubic(const double *p, int p_stride, double x,
                              double y) {
@@ -1080,9 +1087,8 @@
   assert(xi > 0);
 
   const double *prate = &interp_rgrid_curv[rcat][(xi - 1)];
-  *rate_f = interp_cubic(prate, xo);
   const double *pdist = &interp_dgrid_curv[dcat][(xi - 1)];
-  *distbysse_f = interp_cubic(pdist, xo);
+  av1_interp_cubic_rate_dist(prate, pdist, xo, rate_f, distbysse_f);
 }
 
 static void get_entropy_contexts_plane(BLOCK_SIZE plane_bsize,
diff --git a/av1/encoder/x86/model_rd_sse4.c b/av1/encoder/x86/model_rd_sse4.c
new file mode 100644
index 0000000..48bc580
--- /dev/null
+++ b/av1/encoder/x86/model_rd_sse4.c
@@ -0,0 +1,64 @@
+/*
+ * Copyright (c) 2026, Alliance for Open Media. All rights reserved.
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include <smmintrin.h>
+
+#include "config/av1_rtcd.h"
+
+void av1_interp_cubic_rate_dist_sse4_1(const double *p1, const double *p2,
+                                       double x, double *const rate_f,
+                                       double *const distbysse_f) {
+  const __m128d half = _mm_set1_pd(0.5);
+  const __m128d two = _mm_set1_pd(2.0);
+  const __m128d three = _mm_set1_pd(3.0);
+  const __m128d four = _mm_set1_pd(4.0);
+  const __m128d five = _mm_set1_pd(5.0);
+
+  const __m128d reg_x = _mm_set1_pd(x);
+  const __m128d reg_p0 = _mm_set_pd(p2[0], p1[0]);
+  const __m128d reg_p1 = _mm_set_pd(p2[1], p1[1]);
+  const __m128d reg_p2 = _mm_set_pd(p2[2], p1[2]);
+  const __m128d reg_p3 = _mm_set_pd(p2[3], p1[3]);
+
+  // To ensure that results are bit-identical to the C code, we need to perform
+  // exactly the same sequence of operations here as in the C code.
+  // reg_res_0 = x * (3.0 * (p[1] - p[2]) + p[3] - p[0])
+  __m128d reg_res_0 = _mm_sub_pd(reg_p1, reg_p2);
+  reg_res_0 = _mm_mul_pd(three, reg_res_0);
+  reg_res_0 = _mm_add_pd(reg_res_0, reg_p3);
+  reg_res_0 = _mm_sub_pd(reg_res_0, reg_p0);
+  reg_res_0 = _mm_mul_pd(reg_x, reg_res_0);
+
+  // reg_res_1 = 2.0 * p[0] - 5.0 * p[1] + 4.0 * p[2]- p[3]
+  const __m128d regp0_x_2 = _mm_mul_pd(two, reg_p0);
+  const __m128d regp1_x_5 = _mm_mul_pd(five, reg_p1);
+  const __m128d regp2_x_4 = _mm_mul_pd(four, reg_p2);
+  __m128d reg_res_1 = _mm_sub_pd(regp0_x_2, regp1_x_5);
+  reg_res_1 = _mm_add_pd(reg_res_1, regp2_x_4);
+  reg_res_1 = _mm_sub_pd(reg_res_1, reg_p3);
+
+  // reg_res_2 = x * (reg_res_1 + reg_res_0)
+  __m128d reg_res_2 = _mm_add_pd(reg_res_1, reg_res_0);
+  reg_res_2 = _mm_mul_pd(reg_x, reg_res_2);
+
+  // reg_res_3 = p[2] - p[0] + reg_res_2
+  __m128d reg_res_3 = _mm_sub_pd(reg_p2, reg_p0);
+  reg_res_3 = _mm_add_pd(reg_res_3, reg_res_2);
+
+  // reg_res_4 = p[1] + 0.5 * x * reg_res_3
+  __m128d reg_res_4 = _mm_mul_pd(_mm_mul_pd(half, reg_x), reg_res_3);
+  reg_res_4 = _mm_add_pd(reg_p1, reg_res_4);
+
+  double result[2];
+  _mm_storeu_pd(result, reg_res_4);
+  *rate_f = result[0];
+  *distbysse_f = result[1];
+}
diff --git a/test/model_rd_test.cc b/test/model_rd_test.cc
new file mode 100644
index 0000000..910f20a
--- /dev/null
+++ b/test/model_rd_test.cc
@@ -0,0 +1,114 @@
+/*
+ * Copyright (c) 2026, Alliance for Open Media. All rights reserved.
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include <cstdlib>
+
+#include "gtest/gtest.h"
+#include "config/av1_rtcd.h"
+
+#include "test/acm_random.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+
+using libaom_test::ACMRandom;
+
+namespace {
+
+using InterpCubicRateDistFunc = void (*)(const double *p1, const double *p2,
+                                         double x, double *const rate_f,
+                                         double *const distbysse_f);
+
+using InterpCubicTestParam = std::tuple<const InterpCubicRateDistFunc>;
+
+class InterpCubicTest : public ::testing::TestWithParam<InterpCubicTestParam> {
+ public:
+  double generate_random_double(double min, double max) {
+    return min + (static_cast<double>(rnd_.Rand31()) / ((1U << 31) - 1)) *
+                     (max - min);
+  }
+  void SetUp() override { target_func_ = GET_PARAM(0); }
+  void TearDown() override {}
+  void CheckOutput();
+  void SpeedTest();
+
+ protected:
+  InterpCubicRateDistFunc target_func_;
+
+ private:
+  libaom_test::ACMRandom rnd_;
+};
+GTEST_ALLOW_UNINSTANTIATED_PARAMETERIZED_TEST(InterpCubicTest);
+
+void InterpCubicTest::CheckOutput() {
+  double p1[4], p2[4];
+  double rate_f_ref, rate_f_mod, distbysse_f_ref, distbysse_f_mod;
+  const int knum_iter = 10000;
+  for (int iter = 0; iter < knum_iter; iter++) {
+    for (int i = 0; i < 4; i++) {
+      p1[i] = generate_random_double(0.0000, 4096.000000);
+      p2[i] = generate_random_double(0.0000, 16.0000);
+    }
+    double x = generate_random_double(0.0000, 1.0000);
+
+    av1_interp_cubic_rate_dist_c(p1, p2, x, &rate_f_ref, &distbysse_f_ref);
+    target_func_(p1, p2, x, &rate_f_mod, &distbysse_f_mod);
+    EXPECT_EQ(rate_f_ref, rate_f_mod) << "Error: rate_f value mismatch";
+    EXPECT_EQ(distbysse_f_ref, distbysse_f_mod)
+        << "Error: distbysse_f value mismatch";
+  }
+}
+
+void InterpCubicTest::SpeedTest() {
+  double p1[4], p2[4];
+  double rate_f_ref, rate_f_mod, distbysse_f_ref, distbysse_f_mod;
+
+  for (int i = 0; i < 4; i++) {
+    p1[i] = generate_random_double(0.0000, 4096.0000);
+    p2[i] = generate_random_double(0.0000, 16.0000);
+  }
+  double x = generate_random_double(0.0000, 1.0000);
+
+  const int num_iter = 100000000;
+
+  aom_usec_timer ref_timer, test_timer;
+  aom_usec_timer_start(&ref_timer);
+  for (int iter = 0; iter < num_iter; iter++) {
+    av1_interp_cubic_rate_dist_c(p1, p2, x, &rate_f_ref, &distbysse_f_ref);
+  }
+  aom_usec_timer_mark(&ref_timer);
+  const int elapsed_time_c =
+      static_cast<int>(aom_usec_timer_elapsed(&ref_timer));
+
+  aom_usec_timer_start(&test_timer);
+  for (int iter = 0; iter < num_iter; iter++) {
+    target_func_(p1, p2, x, &rate_f_mod, &distbysse_f_mod);
+  }
+  aom_usec_timer_mark(&test_timer);
+  const int elapsed_time_simd =
+      static_cast<int>(aom_usec_timer_elapsed(&test_timer));
+
+  printf(
+      " c_time=%d \t simd_time=%d \t "
+      "Scaling=%lf \n",
+      elapsed_time_c, elapsed_time_simd,
+      (static_cast<double>(elapsed_time_c) / elapsed_time_simd));
+}
+
+TEST_P(InterpCubicTest, CheckOutput) { CheckOutput(); }
+
+TEST_P(InterpCubicTest, DISABLED_Speed) { SpeedTest(); }
+
+#if HAVE_SSE4_1
+INSTANTIATE_TEST_SUITE_P(SSE4_1, InterpCubicTest,
+                         ::testing::Values(av1_interp_cubic_rate_dist_sse4_1));
+#endif  // HAVE_SSE4_1
+
+}  // namespace
diff --git a/test/test.cmake b/test/test.cmake
index 6224c54..84ba918 100644
--- a/test/test.cmake
+++ b/test/test.cmake
@@ -232,6 +232,7 @@
               "${AOM_ROOT}/test/masked_variance_test.cc"
               "${AOM_ROOT}/test/metadata_test.cc"
               "${AOM_ROOT}/test/minmax_test.cc"
+              "${AOM_ROOT}/test/model_rd_test.cc"
               "${AOM_ROOT}/test/motion_vector_test.cc"
               "${AOM_ROOT}/test/mv_cost_test.cc"
               "${AOM_ROOT}/test/obmc_sad_test.cc"