diff --git a/examples/cli/main.cpp b/examples/cli/main.cpp
index ceae27b83..51afba9e2 100644
--- a/examples/cli/main.cpp
+++ b/examples/cli/main.cpp
@@ -71,6 +71,7 @@ struct SDParams {
     SDMode mode   = TXT2IMG;
 
     std::string model_path;
+    std::string clip_g_path;
     std::string clip_l_path;
     std::string t5xxl_path;
     std::string diffusion_model_path;
@@ -127,6 +128,7 @@ void print_params(SDParams params) {
     printf("    mode:              %s\n", modes_str[params.mode]);
     printf("    model_path:        %s\n", params.model_path.c_str());
     printf("    wtype:             %s\n", params.wtype < SD_TYPE_COUNT ? sd_type_name(params.wtype) : "unspecified");
+    printf("    clip_g_path:       %s\n", params.clip_g_path.c_str());
     printf("    clip_l_path:       %s\n", params.clip_l_path.c_str());
     printf("    t5xxl_path:        %s\n", params.t5xxl_path.c_str());
     printf("    diffusion_model_path:   %s\n", params.diffusion_model_path.c_str());
@@ -175,6 +177,7 @@ void print_usage(int argc, const char* argv[]) {
     printf("                                     If threads <= 0, then threads will be set to the number of CPU physical cores\n");
     printf("  -m, --model [MODEL]                path to full model\n");
     printf("  --diffusion-model                  path to the standalone diffusion model\n");
+    printf("  --clip_g                           path to the clip-g text encoder\n");
     printf("  --clip_l                           path to the clip-l text encoder\n");
     printf("  --t5xxl                            path to the the t5xxl text encoder.\n");
     printf("  --vae [VAE]                        path to vae\n");
@@ -256,6 +259,12 @@ void parse_args(int argc, const char** argv, SDParams& params) {
                 break;
             }
             params.model_path = argv[i];
+        } else if (arg == "--clip_g") {
+            if (++i >= argc) {
+                invalid_arg = true;
+                break;
+            }
+            params.clip_g_path = argv[i];
         } else if (arg == "--clip_l") {
             if (++i >= argc) {
                 invalid_arg = true;
@@ -764,6 +773,7 @@ int main(int argc, const char* argv[]) {
     }
 
     sd_ctx_t* sd_ctx = new_sd_ctx(params.model_path.c_str(),
+                                  params.clip_g_path.c_str(),
                                   params.clip_l_path.c_str(),
                                   params.t5xxl_path.c_str(),
                                   params.diffusion_model_path.c_str(),
diff --git a/stable-diffusion.cpp b/stable-diffusion.cpp
index 07b59bb8a..87b8e3f3e 100644
--- a/stable-diffusion.cpp
+++ b/stable-diffusion.cpp
@@ -138,6 +138,7 @@ class StableDiffusionGGML {
     }
 
     bool load_from_file(const std::string& model_path,
+                        const std::string& clip_g_path,
                         const std::string& clip_l_path,
                         const std::string& t5xxl_path,
                         const std::string& diffusion_model_path,
@@ -167,7 +168,7 @@ class StableDiffusionGGML {
         for (int device = 0; device < ggml_backend_vk_get_device_count(); ++device) {
             backend = ggml_backend_vk_init(device);
         }
-        if(!backend) {
+        if (!backend) {
             LOG_WARN("Failed to initialize Vulkan backend");
         }
 #endif
@@ -181,7 +182,7 @@ class StableDiffusionGGML {
             backend = ggml_backend_cpu_init();
         }
 #ifdef SD_USE_FLASH_ATTENTION
-#if defined(SD_USE_CUBLAS) || defined(SD_USE_METAL) || defined (SD_USE_SYCL) || defined(SD_USE_VULKAN)
+#if defined(SD_USE_CUBLAS) || defined(SD_USE_METAL) || defined(SD_USE_SYCL) || defined(SD_USE_VULKAN)
         LOG_WARN("Flash Attention not supported with GPU Backend");
 #else
         LOG_INFO("Flash Attention enabled");
@@ -198,24 +199,44 @@ class StableDiffusionGGML {
             }
         }
 
+        if (diffusion_model_path.size() > 0) {
+            LOG_INFO("loading diffusion model from '%s'", diffusion_model_path.c_str());
+            if (!model_loader.init_from_file(diffusion_model_path, "model.diffusion_model.")) {
+                LOG_WARN("loading diffusion model from '%s' failed", diffusion_model_path.c_str());
+            }
+        }
+        version = model_loader.get_sd_version();
+
+        if (clip_g_path.size() > 0) {
+            LOG_INFO("loading clip_g from '%s'", clip_g_path.c_str());
+            std::string prefix = "text_encoders.clip_g.";
+            if (version == VERSION_SD3_2B ) {
+                prefix = "text_encoders.clip_g.transformer.";
+            }
+            if (!model_loader.init_from_file(clip_g_path, prefix)) {
+                LOG_WARN("loading clip_g from '%s' failed", clip_g_path.c_str());
+            }
+        }
+
         if (clip_l_path.size() > 0) {
             LOG_INFO("loading clip_l from '%s'", clip_l_path.c_str());
-            if (!model_loader.init_from_file(clip_l_path, "text_encoders.clip_l.")) {
+            std::string prefix = "text_encoders.clip_l.";
+            if (version == VERSION_SD3_2B ) {
+                prefix = "text_encoders.clip_l.transformer.";
+            }
+            if (!model_loader.init_from_file(clip_l_path, prefix)) {
                 LOG_WARN("loading clip_l from '%s' failed", clip_l_path.c_str());
             }
         }
 
         if (t5xxl_path.size() > 0) {
             LOG_INFO("loading t5xxl from '%s'", t5xxl_path.c_str());
-            if (!model_loader.init_from_file(t5xxl_path, "text_encoders.t5xxl.")) {
-                LOG_WARN("loading t5xxl from '%s' failed", t5xxl_path.c_str());
+            std::string prefix = "text_encoders.t5xxl.";
+            if (version == VERSION_SD3_2B ) {
+                prefix = "text_encoders.t5xxl.transformer.";
             }
-        }
-
-        if (diffusion_model_path.size() > 0) {
-            LOG_INFO("loading diffusion model from '%s'", diffusion_model_path.c_str());
-            if (!model_loader.init_from_file(diffusion_model_path, "model.diffusion_model.")) {
-                LOG_WARN("loading diffusion model from '%s' failed", diffusion_model_path.c_str());
+            if (!model_loader.init_from_file(t5xxl_path, prefix)) {
+                LOG_WARN("loading t5xxl from '%s' failed", t5xxl_path.c_str());
             }
         }
 
@@ -226,7 +247,6 @@ class StableDiffusionGGML {
             }
         }
 
-        version = model_loader.get_sd_version();
         if (version == VERSION_COUNT) {
             LOG_ERROR("get sd version from file failed: '%s'", model_path.c_str());
             return false;
@@ -1007,6 +1027,7 @@ struct sd_ctx_t {
 };
 
 sd_ctx_t* new_sd_ctx(const char* model_path_c_str,
+                     const char* clip_g_path_c_str,
                      const char* clip_l_path_c_str,
                      const char* t5xxl_path_c_str,
                      const char* diffusion_model_path_c_str,
@@ -1031,6 +1052,7 @@ sd_ctx_t* new_sd_ctx(const char* model_path_c_str,
         return NULL;
     }
     std::string model_path(model_path_c_str);
+    std::string clip_g_path(clip_g_path_c_str);
     std::string clip_l_path(clip_l_path_c_str);
     std::string t5xxl_path(t5xxl_path_c_str);
     std::string diffusion_model_path(diffusion_model_path_c_str);
@@ -1051,6 +1073,7 @@ sd_ctx_t* new_sd_ctx(const char* model_path_c_str,
     }
 
     if (!sd_ctx->sd->load_from_file(model_path,
+                                    clip_g_path,
                                     clip_l_path,
                                     t5xxl_path_c_str,
                                     diffusion_model_path,
diff --git a/stable-diffusion.h b/stable-diffusion.h
index 0d4cc1fda..d6b5a855c 100644
--- a/stable-diffusion.h
+++ b/stable-diffusion.h
@@ -123,6 +123,7 @@ typedef struct {
 typedef struct sd_ctx_t sd_ctx_t;
 
 SD_API sd_ctx_t* new_sd_ctx(const char* model_path,
+                            const char* clip_g_path,
                             const char* clip_l_path,
                             const char* t5xxl_path,
                             const char* diffusion_model_path,