diff --git a/llama.cpp b/llama.cpp
index 5360f13ce..2469e5efd 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -12479,17 +12479,20 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n
         }
     } else if (ftype == LLAMA_FTYPE_MOSTLY_IQ1_XS) {
         if (name.find("attn_q.weight") != std::string::npos) {
-            if (qs.model.hparams.n_expert >= 2) new_type = GGML_TYPE_IQ2_S;
+            if (qs.model.hparams.n_expert >= 8) new_type = GGML_TYPE_IQ2_S;
+            else if (qs.model.hparams.n_expert >= 4) new_type = GGML_TYPE_IQ2_XS;
+            else if (qs.model.hparams.n_expert >= 2) new_type = GGML_TYPE_IQ2_XXS;
         }
         if (name.find("attn_k.weight") != std::string::npos) {
             if (qs.model.hparams.n_expert >= 8) new_type = GGML_TYPE_IQ4_XS;
             else if (qs.model.hparams.n_expert >= 4) new_type = GGML_TYPE_IQ3_S;
             else if (qs.model.hparams.n_expert >= 2) new_type = GGML_TYPE_IQ3_XXS;	
+            else if (qs.model.hparams.n_gqa() >= 8) new_type = GGML_TYPE_IQ2_S;
             else if (qs.model.hparams.n_gqa() >= 4) new_type = GGML_TYPE_IQ2_XS;
             else if (qs.model.hparams.n_gqa() >= 2) new_type = GGML_TYPE_IQ2_XXS;
         }
         else if (name.find("attn_v.weight") != std::string::npos) {
-            if (qs.model.hparams.n_expert >= 8) new_type = GGML_TYPE_IQ4_XS;
+            if (qs.model.hparams.n_expert >= 8) new_type = GGML_TYPE_Q5_K;
             else if (qs.model.hparams.n_gqa() >= 8 || qs.model.hparams.n_expert >= 4) new_type = GGML_TYPE_IQ4_XS;
             else if (qs.model.hparams.n_gqa() >= 4 || qs.model.hparams.n_expert >= 2) new_type = GGML_TYPE_IQ3_S;
             else if (qs.model.hparams.n_gqa() >= 2) new_type = GGML_TYPE_IQ3_XXS;