diff --git a/examples/llava/clip-quantize-cli.cpp b/examples/llava/clip-quantize-cli.cpp
deleted file mode 100644
index 566506954..000000000
--- a/examples/llava/clip-quantize-cli.cpp
+++ /dev/null
@@ -1,59 +0,0 @@
-#include "arg.h"
-#include "base64.hpp"
-#include "log.h"
-#include "common.h"
-#include "sampling.h"
-#include "clip.h"
-#include "llava.h"
-#include "llama.h"
-#include "ggml.h"
-
-static void print_usage(int argc, char ** argv) {
-    (void) argc;
-
-    fprintf(stderr, "usage: %s /path/to/ggml-model-f32.gguf /path/to/ggml-model-quantized.gguf type\n", argv[0]);
-    fprintf(stderr, "  type = 2 - q4_0\n");
-    fprintf(stderr, "  type = 3 - q4_1\n");
-    fprintf(stderr, "  type = 6 - q5_0\n");
-    fprintf(stderr, "  type = 7 - q5_1\n");
-    fprintf(stderr, "  type = 8 - q8_0\n");
-}
-
-int main(int argc, char ** argv) {
-    if (argc != 4) {
-        print_usage(argc, argv);
-        return 1;
-    }
-
-    const std::string fname_inp = argv[1];
-    const std::string fname_out = argv[2];
-
-    const int itype = atoi(argv[3]);
-
-    const int64_t t_main_start_us = ggml_time_us();
-
-    int64_t t_quantize_us = 0;
-
-    // load the model
-    {
-        const int64_t t_start_us = ggml_time_us();
-
-        if (!clip_model_quantize(fname_inp.c_str(), fname_out.c_str(), itype)) {
-            fprintf(stderr, "%s: failed to quantize model from '%s'\n", __func__, fname_inp.c_str());
-            return 1;
-        }
-
-        t_quantize_us = ggml_time_us() - t_start_us;
-    }
-
-    // report timing
-    {
-        const int64_t t_main_end_us = ggml_time_us();
-
-        printf("\n");
-        printf("%s: quantize time = %8.2f ms\n", __func__, t_quantize_us / 1000.0f);
-        printf("%s:    total time = %8.2f ms\n", __func__, (t_main_end_us - t_main_start_us) / 1000.0f);
-    }
-
-    return 0;
-}
diff --git a/examples/llava/quantclip.cpp b/examples/llava/quantclip.cpp
index 65a3a7c44..7173913a7 100644
--- a/examples/llava/quantclip.cpp
+++ b/examples/llava/quantclip.cpp
@@ -16,7 +16,7 @@ int main(int argc, char ** argv) {
 
      if (argc != 3 && argc != 4) {
         fprintf(stderr, "usage: %s mmproj-f16.gguf output-mmproj-quantized.gguf TYPE\n", argv[0]);
-        printf("\nGGML_TYPE_Q4_0    = 2\nGGML_TYPE_Q4_1    = 3\nGGML_TYPE_Q5_0    = 6\nGGML_TYPE_Q5_1    = 7\nGGML_TYPE_Q8_0    = 8\n");
+        printf("\nGGML_TYPE_F16    = 1\nGGML_TYPE_Q4_0    = 2\nGGML_TYPE_Q4_1    = 3\nGGML_TYPE_Q5_0    = 6\nGGML_TYPE_Q5_1    = 7\nGGML_TYPE_Q8_0    = 8\n");
 
         return 1;
     }