diff --git a/src/llama-model.cpp b/src/llama-model.cpp
--- a/src/llama-model.cpp
+++ b/src/llama-model.cpp
@@ -15,8 +15,6 @@
 #include "llama-memory-hybrid-iswa.h"
 #include "llama-memory-recurrent.h"
 
-#include "models/models.h"
-
 #include "ggml.h"
 #include "ggml-cpp.h"
 
@@ -35,6 +33,8 @@
 #include <string>
 #include <vector>
 
+#include "models/models.h"
+
 static llama_model * llama_model_mapping(llm_arch arch, const llama_model_params & params) {
     switch (arch) {
         case LLM_ARCH_LLAMA:
diff --git a/src/models/chameleon.cpp b/src/models/chameleon.cpp
--- a/src/models/chameleon.cpp
+++ b/src/models/chameleon.cpp
@@ -1,5 +1,5 @@
+#include <float.h>
 #include "models.h"
-#include <float.h>
 
 void llama_model_chameleon::load_arch_hparams(llama_model_loader & ml) {
     ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
diff --git a/src/models/deepseek32.cpp b/src/models/deepseek32.cpp
--- a/src/models/deepseek32.cpp
+++ b/src/models/deepseek32.cpp
@@ -1,7 +1,6 @@
-#include "models.h"
-
 #include "llama-kv-cache.h"
 #include "llama-kv-cache-dsa.h"
+#include "models.h"
 
 void llama_model_deepseek32::load_arch_hparams(llama_model_loader & ml) {
     ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,     hparams.n_ff_exp);
diff --git a/src/models/delta-net-base.cpp b/src/models/delta-net-base.cpp
--- a/src/models/delta-net-base.cpp
+++ b/src/models/delta-net-base.cpp
@@ -1,7 +1,6 @@
-#include "models.h"
-
 #include "llama-impl.h"
 #include "llama-memory-recurrent.h"
+#include "models.h"
 
 // utility to get one slice from the third dimension
 // input dim:  [x, y, c, b]
diff --git a/src/models/granite.cpp b/src/models/granite.cpp
--- a/src/models/granite.cpp
+++ b/src/models/granite.cpp
@@ -1,7 +1,6 @@
+#include <sstream>
 #include "models.h"
 
-#include <sstream>
-
 void llama_model_granite::load_arch_hparams(llama_model_loader & ml) {
     ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
     ml.get_key(LLM_KV_LOGIT_SCALE,                 hparams.f_logit_scale);
diff --git a/src/models/kimi-linear.cpp b/src/models/kimi-linear.cpp
--- a/src/models/kimi-linear.cpp
+++ b/src/models/kimi-linear.cpp
@@ -1,5 +1,5 @@
+#include "llama-memory-recurrent.h"
 #include "models.h"
-#include "llama-memory-recurrent.h"
 
 void llama_model_kimi_linear::load_arch_hparams(llama_model_loader & ml) {
     ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
diff --git a/src/models/lfm2.cpp b/src/models/lfm2.cpp
--- a/src/models/lfm2.cpp
+++ b/src/models/lfm2.cpp
@@ -1,6 +1,6 @@
-#include "models.h"
 #include "../llama-memory-hybrid-iswa.h"
 #include "../llama-memory-hybrid.h"
+#include "models.h"
 
 void llama_model_lfm2::load_arch_hparams(llama_model_loader & ml) {
     ml.get_key(LLM_KV_SHORTCONV_L_CACHE,           hparams.n_shortconv_l_cache);
diff --git a/src/models/lfm2moe.cpp b/src/models/lfm2moe.cpp
--- a/src/models/lfm2moe.cpp
+++ b/src/models/lfm2moe.cpp
@@ -1,6 +1,6 @@
-#include "models.h"
 #include "../llama-memory-hybrid-iswa.h"
 #include "../llama-memory-hybrid.h"
+#include "models.h"
 
 void llama_model_lfm2moe::load_arch_hparams(llama_model_loader & ml) {
     ml.get_key(LLM_KV_SHORTCONV_L_CACHE,           hparams.n_shortconv_l_cache);
diff --git a/src/models/mamba-base.cpp b/src/models/mamba-base.cpp
--- a/src/models/mamba-base.cpp
+++ b/src/models/mamba-base.cpp
@@ -1,7 +1,6 @@
+#include "llama-memory-recurrent.h"
 #include "models.h"
 
-#include "llama-memory-recurrent.h"
-
 llm_build_mamba_base::llm_build_mamba_base(const llm_graph_params & params) : llm_graph_context(params) {}
 
 ggml_tensor * llm_build_mamba_base::build_mamba_layer(llm_graph_input_rs * inp,
diff --git a/src/models/plamo2.cpp b/src/models/plamo2.cpp
--- a/src/models/plamo2.cpp
+++ b/src/models/plamo2.cpp
@@ -1,5 +1,5 @@
+#include "llama-memory-recurrent.h"
 #include "models.h"
-#include "llama-memory-recurrent.h"
 
 void llama_model_plamo2::load_arch_hparams(llama_model_loader & ml) {
     ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
diff --git a/src/models/qwen35.cpp b/src/models/qwen35.cpp
--- a/src/models/qwen35.cpp
+++ b/src/models/qwen35.cpp
@@ -1,5 +1,5 @@
+#include "llama-memory-recurrent.h"
 #include "models.h"
-#include "llama-memory-recurrent.h"
 
 void llama_model_qwen35::load_arch_hparams(llama_model_loader & ml) {
     ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,       hparams.f_norm_rms_eps);
diff --git a/src/models/qwen35moe.cpp b/src/models/qwen35moe.cpp
--- a/src/models/qwen35moe.cpp
+++ b/src/models/qwen35moe.cpp
@@ -1,5 +1,5 @@
+#include "llama-memory-recurrent.h"
 #include "models.h"
-#include "llama-memory-recurrent.h"
 
 void llama_model_qwen35moe::load_arch_hparams(llama_model_loader & ml) {
     ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,        hparams.n_ff_exp, false);
diff --git a/src/models/qwen3next.cpp b/src/models/qwen3next.cpp
--- a/src/models/qwen3next.cpp
+++ b/src/models/qwen3next.cpp
@@ -1,5 +1,5 @@
+#include "llama-memory-recurrent.h"
 #include "models.h"
-#include "llama-memory-recurrent.h"
 
 void llama_model_qwen3next::load_arch_hparams(llama_model_loader & ml) {
     ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,        hparams.n_ff_exp, false);
diff --git a/src/models/rwkv6-base.cpp b/src/models/rwkv6-base.cpp
--- a/src/models/rwkv6-base.cpp
+++ b/src/models/rwkv6-base.cpp
@@ -1,7 +1,6 @@
+#include "llama-memory-recurrent.h"
 #include "models.h"
 
-#include "llama-memory-recurrent.h"
-
 llm_build_rwkv6_base::llm_build_rwkv6_base(const llama_model & model, const llm_graph_params & params) :
     llm_graph_context(params),
     model(model) {}
diff --git a/src/models/rwkv7-base.cpp b/src/models/rwkv7-base.cpp
--- a/src/models/rwkv7-base.cpp
+++ b/src/models/rwkv7-base.cpp
@@ -1,7 +1,6 @@
+#include "llama-memory-recurrent.h"
 #include "models.h"
 
-#include "llama-memory-recurrent.h"
-
 llm_build_rwkv7_base::llm_build_rwkv7_base(const llama_model & model, const llm_graph_params & params) :
     llm_graph_context(params),
     model(model) {}