diff --git a/modules/core/include/opencv2/core/hal/intrin_rvv_scalable.hpp b/modules/core/include/opencv2/core/hal/intrin_rvv_scalable.hpp
index 67c9d12741..ca82d140f8 100644
--- a/modules/core/include/opencv2/core/hal/intrin_rvv_scalable.hpp
+++ b/modules/core/include/opencv2/core/hal/intrin_rvv_scalable.hpp
@@ -2145,28 +2145,29 @@ inline v_float64 v_dotprod_expand_fast(const v_int32& a, const v_int32& b, const
 { return v_add(v_dotprod_expand_fast(a, b) , c); }
 #endif
 
-// TODO: only 128 bit now.
 inline v_float32 v_matmul(const v_float32& v, const v_float32& mat0,
                             const v_float32& mat1, const v_float32& mat2,
                             const v_float32& mat3)
 {
-    vfloat32m2_t res;
-    res = __riscv_vfmul_vf_f32m2(mat0, v_extract_n(v, 0), VTraits<v_float32>::vlanes());
-    res = __riscv_vfmacc_vf_f32m2(res, v_extract_n(v, 1), mat1, VTraits<v_float32>::vlanes());
-    res = __riscv_vfmacc_vf_f32m2(res, v_extract_n(v, 2), mat2, VTraits<v_float32>::vlanes());
-    res = __riscv_vfmacc_vf_f32m2(res, v_extract_n(v, 3), mat3, VTraits<v_float32>::vlanes());
-    return res;
+    const int vl = VTraits<v_float32>::vlanes();
+    vuint32m2_t idx = __riscv_vand(__riscv_vid_v_u32m2(vl), 0xfffffffc, vl);
+    v_float32 v0 = __riscv_vrgather(v, idx, vl);
+    v_float32 v1 = __riscv_vrgather(v, __riscv_vadd(idx, 1, vl), vl);
+    v_float32 v2 = __riscv_vrgather(v, __riscv_vadd(idx, 2, vl), vl);
+    v_float32 v3 = __riscv_vrgather(v, __riscv_vadd(idx, 3, vl), vl);
+    return v_fma(v0, mat0, v_fma(v1, mat1, v_fma(v2, mat2, v_mul(v3, mat3))));
 }
 
-// TODO: only 128 bit now.
 inline v_float32 v_matmuladd(const v_float32& v, const v_float32& mat0,
                                const v_float32& mat1, const v_float32& mat2,
                                const v_float32& a)
 {
-    vfloat32m2_t res = __riscv_vfmul_vf_f32m2(mat0, v_extract_n(v,0), VTraits<v_float32>::vlanes());
-    res = __riscv_vfmacc_vf_f32m2(res, v_extract_n(v,1), mat1, VTraits<v_float32>::vlanes());
-    res = __riscv_vfmacc_vf_f32m2(res, v_extract_n(v,2), mat2, VTraits<v_float32>::vlanes());
-    return __riscv_vfadd(res, a, VTraits<v_float32>::vlanes());
+    const int vl = VTraits<v_float32>::vlanes();
+    vuint32m2_t idx = __riscv_vand(__riscv_vid_v_u32m2(vl), 0xfffffffc, vl);
+    v_float32 v0 = __riscv_vrgather(v, idx, vl);
+    v_float32 v1 = __riscv_vrgather(v, __riscv_vadd(idx, 1, vl), vl);
+    v_float32 v2 = __riscv_vrgather(v, __riscv_vadd(idx, 2, vl), vl);
+    return v_fma(v0, mat0, v_fma(v1, mat1, v_fma(v2, mat2, a)));
 }
 
 inline void v_cleanup() {}
diff --git a/modules/core/src/matmul.simd.hpp b/modules/core/src/matmul.simd.hpp
index f089c90a01..5645617dd1 100644
--- a/modules/core/src/matmul.simd.hpp
+++ b/modules/core/src/matmul.simd.hpp
@@ -1596,8 +1596,7 @@ static void
 transform_32f( const float* src, float* dst, const float* m, int len, int scn, int dcn )
 {
 // Disabled for RISC-V Vector (scalable), because of:
-// 1. v_matmuladd for RVV is 128-bit only but not scalable, this will fail the test `Core_Transform.accuracy`.
-// 2. Both gcc and clang can autovectorize this, with better performance than using Universal intrinsic.
+// 1. Both gcc and clang can autovectorize this, with better performance than using Universal intrinsic.
 #if (CV_SIMD || CV_SIMD_SCALABLE) && !defined(__aarch64__) && !defined(_M_ARM64) && !defined(_M_ARM64EC) && !(CV_TRY_RVV && CV_RVV)
     int x = 0;
     if( scn == 3 && dcn == 3 )
diff --git a/modules/core/test/test_intrin_utils.hpp b/modules/core/test/test_intrin_utils.hpp
index 8c2c69c79f..2a0ce37611 100644
--- a/modules/core/test/test_intrin_utils.hpp
+++ b/modules/core/test/test_intrin_utils.hpp
@@ -1519,9 +1519,8 @@ template<typename R> struct TheTest
         R v = dataV, a = dataA, b = dataB, c = dataC, d = dataD;
 
         Data<R> res = v_matmul(v, a, b, c, d);
-        // for (int i = 0; i < VTraits<R>::vlanes(); i += 4)
-        // {
-            int i = 0;
+        for (int i = 0; i < VTraits<R>::vlanes(); i += 4)
+        {
             for (int j = i; j < i + 4; ++j)
             {
                 SCOPED_TRACE(cv::format("i=%d j=%d", i, j));
@@ -1531,12 +1530,11 @@ template<typename R> struct TheTest
                              + dataV[i + 3] * dataD[j];
                 EXPECT_COMPARE_EQ(val, res[j]);
             }
-        // }
+        }
 
         Data<R> resAdd = v_matmuladd(v, a, b, c, d);
-        // for (int i = 0; i < VTraits<R>::vlanes(); i += 4)
-        // {
-            i = 0;
+        for (int i = 0; i < VTraits<R>::vlanes(); i += 4)
+        {
             for (int j = i; j < i + 4; ++j)
             {
                 SCOPED_TRACE(cv::format("i=%d j=%d", i, j));
@@ -1546,7 +1544,7 @@ template<typename R> struct TheTest
                              + dataD[j];
                 EXPECT_COMPARE_EQ(val, resAdd[j]);
             }
-        // }
+        }
         return *this;
     }