diff --git a/hwy/contrib/dot/dot-inl.h b/hwy/contrib/dot/dot-inl.h
index e48796aaa7..10fd897918 100644
--- a/hwy/contrib/dot/dot-inl.h
+++ b/hwy/contrib/dot/dot-inl.h
@@ -174,9 +174,17 @@ struct Dot {
         (kAssumptions & kMultipleOfVector) != 0;
     constexpr bool kIsPaddedToVector = (kAssumptions & kPaddedToVector) != 0;
 
-    // Won't be able to do a full vector load without padding => scalar loop.
+    // Won't be able to do a full vector load without padding. Use a scalar
+    // loop under Clang. GCC has very suboptimal codegen for scalar BF16->float
+    // conversions, so use vector ops with LoadN instead.
+    // TODO: https://github.com/google/highway/pull/2703
     if (!kIsAtLeastOneVector && !kIsMultipleOfVector && !kIsPaddedToVector &&
         HWY_UNLIKELY(num_elements < NF)) {
+#if HWY_COMPILER_GCC_ACTUAL
+      const VF a = LoadN(df, pa, num_elements);
+      const VF b = PromoteTo(df, LoadN(dbfh, pb, num_elements));
+      return ReduceSum(df, Mul(a, b));
+#else
       // Only 2x unroll to avoid excessive code size.
       float sum0 = 0.0f;
       float sum1 = 0.0f;
@@ -189,6 +197,7 @@ struct Dot {
         sum1 += pa[i] * ConvertScalarTo<float>(pb[i]);
       }
       return sum0 + sum1;
+#endif
     }
 
     // Compiler doesn't make independent sum* accumulators, so unroll manually.
@@ -279,9 +288,19 @@ struct Dot {
         (kAssumptions & kMultipleOfVector) != 0;
     constexpr bool kIsPaddedToVector = (kAssumptions & kPaddedToVector) != 0;
 
-    // Won't be able to do a full vector load without padding => scalar loop.
+    // Won't be able to do a full vector load without padding. Use a scalar
+    // loop under Clang. GCC has very suboptimal codegen for scalar BF16->float
+    // conversions, so use vector ops with LoadN instead.
+    // TODO: https://github.com/google/highway/pull/2703
     if (!kIsAtLeastOneVector && !kIsMultipleOfVector && !kIsPaddedToVector &&
         HWY_UNLIKELY(num_elements < N)) {
+#if HWY_COMPILER_GCC_ACTUAL
+      const auto a = LoadN(d, pa, num_elements);
+      const auto b = LoadN(d, pb, num_elements);
+      V sum1 = Zero(df32);
+      V sum0 = ReorderWidenMulAccumulate(df32, a, b, Zero(df32), sum1);
+      return ReduceSum(df32, Add(sum0, sum1));
+#else
       float sum0 = 0.0f;  // Only 2x unroll to avoid excessive code size for..
       float sum1 = 0.0f;  // this unlikely(?) case.
       for (; i + 2 <= num_elements; i += 2) {
@@ -292,6 +311,7 @@ struct Dot {
         sum1 += F32FromBF16(pa[i]) * F32FromBF16(pb[i]);
       }
       return sum0 + sum1;
+#endif
     }
 
     // See comment in the other Compute() overload. Unroll 2x, but we need