diff --git a/ggml-opencl.cpp b/ggml-opencl.cpp
index 79b20f8d2..d2180d9c2 100644
--- a/ggml-opencl.cpp
+++ b/ggml-opencl.cpp
@@ -820,7 +820,7 @@ static void ggml_cl_mul_mat_q_f32(const ggml_tensor * src0, const ggml_tensor *
                 CL_CHECK(clSetKernelArg(*to_fp32_cl, 0, sizeof(cl_mem), &d_Q), "clSetKernelArg");
                 CL_CHECK(clSetKernelArg(*to_fp32_cl, 1, sizeof(cl_mem), &d_X), "clSetKernelArg");
                 CL_CHECK(clFinish(queue), "clFinish");
-                CL_CHECK(clEnqueueNDRangeKernel(queue, *to_fp32_cl, 1, NULL, &global, &local, 0, NULL, NULL), "clEnqueueNDRangeKernel");
+                CL_CHECK(clEnqueueNDRangeKernel(queue, *to_fp32_cl, 1, NULL, &global, type == GGML_TYPE_F16 ? NULL : &local, 0, NULL, NULL), "clEnqueueNDRangeKernel");
 
                 // copy src1 to device
                 CL_CHECK(ggml_cl_h2d_tensor_2d(queue, d_Y, 0, src1, i03, i02, NULL), "ggml_cl_h2d_tensor_2d");