diff --git a/llama.cpp b/llama.cpp
index d7986b706..dcbe1d4e8 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -4301,9 +4301,9 @@ static bool llm_load_tensors(
                     {
                         model.output_norm   = ml.create_tensor(ctx_output,       tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
                         model.output_norm_b = ml.create_tensor(ctx_output,       tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd});
-                        if (gguf_find_tensor(ml.ctx_gguf, tn(LLM_TENSOR_OUTPUT, "weight").c_str()) >= 0) {
-                            model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT,     "weight"), {n_embd, n_vocab});
-                        } else {
+
+                        model.output        = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab});
+                        if (!model.output) {
                             model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); // needs to be on GPU
                             ml.n_created--; // artificial tensor
                             ml.size_data += ggml_nbytes(model.output);
@@ -4508,9 +4508,8 @@ static bool llm_load_tensors(
                         model.output_norm   = ml.create_tensor(ctx_output,       tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
                         model.output_norm_b = ml.create_tensor(ctx_output,       tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, false);
 
-                        if (gguf_find_tensor(ml.ctx_gguf, tn(LLM_TENSOR_OUTPUT, "weight").c_str()) >= 0) {
-                            model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT,     "weight"), {n_embd, n_vocab});
-                        } else {
+                        model.output        = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab});
+                        if (!model.output) {
                             model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); // needs to be on GPU
                             ml.n_created--; // artificial tensor
                             ml.size_data += ggml_nbytes(model.output);