diff --git a/ggml.c b/ggml.c
index fad137ad6..09299914e 100644
--- a/ggml.c
+++ b/ggml.c
@@ -623,6 +623,17 @@ static inline __m256 sum_i16_pairs_float(const __m128i xh, const __m128i xl) {
     return _mm256_cvtepi32_ps(summed_pairs);
 }
 
+static inline __m256 mul_sum_us8_pairs_float(const __m256i ax, const __m256i sy) {
+    const __m128i axl = _mm256_castsi256_si128(ax);
+    const __m128i axh = _mm256_extractf128_si256(ax, 1);
+    const __m128i syl = _mm256_castsi256_si128(sy);
+    const __m128i syh = _mm256_extractf128_si256(sy, 1);
+    // Perform multiplication and create 16-bit values
+    const __m128i dotl = _mm_maddubs_epi16(axl, syl);
+    const __m128i doth = _mm_maddubs_epi16(axh, syh);
+    return sum_i16_pairs_float(doth, dotl);
+}
+
 // multiply int8_t, add results pairwise twice and return as float vector
 static inline __m256 mul_sum_i8_pairs_float(const __m256i x, const __m256i y) {
     const __m128i xl = _mm256_castsi256_si128(x);