diff --git a/include/llama.h b/include/llama.h
index 5d4afe9bf..a0d5ba5dd 100644
--- a/include/llama.h
+++ b/include/llama.h
@@ -116,12 +116,6 @@ extern "C" {
         LLAMA_ROPE_TYPE_VISION = GGML_ROPE_TYPE_VISION,
     };
 
-    enum llama_expert_gating_func_type {
-        LLAMA_EXPERT_GATING_FUNC_TYPE_NONE    = 0,
-        LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX = 1,
-        LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID = 2,
-    };
-
     enum llama_token_type { //TODO: remove, required until per token attributes are available from GGUF file
         LLAMA_TOKEN_TYPE_UNDEFINED    = 0,
         LLAMA_TOKEN_TYPE_NORMAL       = 1,
diff --git a/src/llama-hparams.h b/src/llama-hparams.h
index 5a5cf0258..a29f20ec4 100644
--- a/src/llama-hparams.h
+++ b/src/llama-hparams.h
@@ -8,6 +8,12 @@
 #define LLAMA_MAX_LAYERS  512
 #define LLAMA_MAX_EXPERTS 256  // DeepSeekV3
 
+enum llama_expert_gating_func_type {
+    LLAMA_EXPERT_GATING_FUNC_TYPE_NONE    = 0,
+    LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX = 1,
+    LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID = 2,
+};
+
 struct llama_hparams_posnet {
     uint32_t n_embd;
     uint32_t n_layer;