diff --git a/python/test/unit/language/test_matmul.py b/python/test/unit/language/test_matmul.py
index 02f7690f4e..bcb0077f6e 100644
--- a/python/test/unit/language/test_matmul.py
+++ b/python/test/unit/language/test_matmul.py
@@ -1298,6 +1298,8 @@ def create_operand(dtype: str, size0: int, size1: int, k_dim: int, transpose: bo
         kernel_kwargs["matrix_instr_nonkdim"] = nonKDim
     if is_xpu() and (128, 256, 256) == (BLOCK_M, BLOCK_N, BLOCK_K) and not CONST_SCALE and not PACK_B_ALONG_K:
         kernel_kwargs["num_warps"] = 8
+    if is_xpu():
+        kernel_kwargs["grf_mode"] = "large"
     out = mxfp8_mxfp4_matmul[grid](a, b, output, a_scale, b_scale, M, N, K, stride_scale, a.stride(0), a.stride(1),
                                    b.stride(0), b.stride(1), output.stride(0), output.stride(1), not CONST_SCALE,
                                    dtype_converter[A_DATA_TYPE], dtype_converter[B_DATA_TYPE], BLOCK_M, BLOCK_N,
diff --git a/scripts/test-triton.sh b/scripts/test-triton.sh
index d9ba1f088c..0baf11db7d 100755
--- a/scripts/test-triton.sh
+++ b/scripts/test-triton.sh
@@ -731,9 +731,9 @@ run_triton_kernels_tests() {
     # FIXME: reconsider in the future
     max_procs=4
   fi
-
+  # skipping mxfp, they are part of mxfp_tests suite
   TRITON_TEST_SUITE=triton_kernels \
-    run_pytest_command -vvv -n $max_procs --device xpu .
+    run_pytest_command -vvv -n $max_procs --device xpu . -k 'not test_mxfp'
 }
 
 test_triton() {