diff --git a/contrib/llama.mak b/contrib/llama.mak
deleted file mode 100644
index d0c819d..0000000
--- a/contrib/llama.mak
+++ /dev/null
@@ -1,256 +0,0 @@
-# llama.cpp server and DLL build (CPU or Vulkan inference)
-#
-# The official llama.cpp build is CMake. This makefile is a replacement
-# build system that produces llama.dll (with the correct exports) and
-# llama-server.exe using w64dk. No source changes required.
-#
-# The DLL exports the public API and no more, and is readily usable as a
-# component in another project (game engine, etc.). The server EXE is
-# fully functional on Windows 7 or later. It is not linked against the
-# DLL because that's not useful.
-#
-# Invoke this makefile in the llama.cpp source tree:
-#
-#   $ make -j$(nproc) -f path/to/w64devkit/contrib/llama.mak
-#
-# To build with Vulkan GPU support, set VULKAN_SDK:
-#
-#   $ make ... VULKAN_SDK=C:/VulkanSDK/1.4.328.1
-#
-# Incremental builds are unsupported, so clean rebuild after pulling. It
-# was last tested at b8565, and an update will inevitably break it.
-
-CC       = gcc
-CXX      = g++
-CROSS    =
-CPPFLAGS = -w -O2 -march=x86-64-v3
-LDFLAGS  = -s
-HOST_CXX = c++
-
-.SUFFIXES: .c .cpp .o
-
-# Parse GGML version from ggml/CMakeLists.txt
-ggml_major := $(shell \
-  sed -n 's/^set(GGML_VERSION_MAJOR \(.*\))/\1/p' ggml/CMakeLists.txt \
-)
-ggml_minor := $(shell \
-  sed -n 's/^set(GGML_VERSION_MINOR \(.*\))/\1/p' ggml/CMakeLists.txt \
-)
-ggml_patch := $(shell \
-  sed -n 's/^set(GGML_VERSION_PATCH \(.*\))/\1/p' ggml/CMakeLists.txt \
-)
-ggml_version = $(ggml_major).$(ggml_minor).$(ggml_patch)
-ggml_commit := $(shell git -C ggml rev-parse --short HEAD || echo unknown)
-
-def = \
-  -DGGML_COMMIT='"$(ggml_commit)"' \
-  -DGGML_USE_CPU \
-  -DGGML_VERSION='"$(ggml_version)"' \
-  -DLLAMA_BUILD_UI \
-  -DLLAMA_USE_HTTPLIB
-inc = \
-  -I. \
-  -Icommon \
-  -Iggml/include \
-  -Iggml/src \
-  -Iggml/src/ggml-cpu \
-  -Iinclude \
-  -Isrc \
-  -Itools/mtmd \
-  -Itools/ui \
-  -Ivendor \
-  -Ivendor/cpp-httplib
-lib =
-
-ui_dist = build/tools/ui/dist
-ui_stamp = $(ui_dist)/.llama.mak.stamp
-ui_assets = index.html bundle.js bundle.css loading.html
-ui_url = https://huggingface.co/buckets/ggml-org/llama-ui/resolve/latest
-
-vk_build_dir = vk-shaders
-vk_shaders_gen = $(vk_build_dir)/vulkan-shaders-gen.exe
-vk_shader_gen_src = $(wildcard \
-  ggml/src/ggml-vulkan/vulkan-shaders/*.cpp \
-)
-vk_shader_gen_obj = $(patsubst \
-  ggml/src/ggml-vulkan/vulkan-shaders/%.cpp, \
-  $(vk_build_dir)/%.cpp.gen.o, \
-  $(vk_shader_gen_src) \
-)
-
-dll = \
-  $(addsuffix .o,$(wildcard \
-      ggml/src/*.c \
-      ggml/src/*.cpp \
-      ggml/src/ggml-cpu/*.c \
-      ggml/src/ggml-cpu/*.cpp \
-      ggml/src/ggml-cpu/arch/x86/*.c \
-      ggml/src/ggml-cpu/arch/x86/*.cpp \
-      src/*.cpp \
-      src/models/*.cpp \
-   ))
-exe = \
-  tools/mtmd/clip.cpp.o \
-  tools/mtmd/mtmd-audio.cpp.o \
-  tools/mtmd/mtmd-helper.cpp.o \
-  tools/mtmd/mtmd-image.cpp.o \
-  tools/mtmd/mtmd.cpp.o \
-  tools/ui/ui.cpp.o \
-  w64dk-build-info.cpp.o \
-  w64dk-license.c.o \
-  $(addsuffix .o,$(wildcard \
-      common/*.cpp \
-      common/jinja/*.cpp \
-      tools/server/*.cpp \
-      tools/mtmd/models/*.cpp \
-      vendor/cpp-httplib/*.cpp \
-   ))
-
-ifdef VULKAN_SDK
-  GLSLC = $(VULKAN_SDK)/Bin/glslc.exe
-  vk_shader_header = $(vk_build_dir)/ggml-vulkan-shaders.hpp
-  vk_shaders = $(wildcard \
-    ggml/src/ggml-vulkan/vulkan-shaders/*.comp \
-  )
-  vk_shader_objs = \
-    $(patsubst %.comp,$(vk_build_dir)/%.comp.cpp.o,$(notdir $(vk_shaders)))
-
-  def += -DGGML_USE_VULKAN
-  inc += -I$(VULKAN_SDK)/Include -I$(vk_build_dir)
-  lib += -L$(VULKAN_SDK)/Lib -lvulkan-1
-  dll += ggml/src/ggml-vulkan/ggml-vulkan.cpp.o $(vk_shader_objs)
-endif
-
-%.c.o: %.c
-	$(CROSS)$(CC) -c -o $@ $(inc) $(def) $(CPPFLAGS) $<
-%.cpp.o: %.cpp
-	$(CROSS)$(CXX) -c -o $@ $(inc) $(def) $(CPPFLAGS) $<
-
-all: llama.dll llama.dll.a llama-server.exe
-
-llama-server.exe: $(exe) $(dll)
-	$(CROSS)$(CXX) $(LDFLAGS) -o $@ $(exe) $(dll) -lws2_32 $(lib)
-
-llama.dll: $(dll) llama.def
-	$(CROSS)$(CXX) -shared $(LDFLAGS) -o $@ $(dll) llama.def $(lib)
-
-llama.dll.a: llama.def
-	$(CROSS)dlltool -l $@ -d $^
-
-clean:
-	rm -f $(dll) $(exe) llama.def llama.dll llama.dll.a llama-server.exe \
-	   tools/ui/index.html.hpp tools/ui/bundle.js.hpp \
-	   tools/ui/bundle.css.hpp tools/ui/loading.html.hpp \
-	   w64dk-build-info.cpp w64dk-license.c vulkan-shaders-gen.exe
-	rm -rf $(vk_build_dir) $(ui_dist)
-
-w64dk-build-info.cpp: common/build-info.cpp.in
-	bnum=`git rev-list --count HEAD 2>/dev/null || printf 0`; \
-	commit=`git rev-parse --short HEAD 2>/dev/null || printf unknown`; \
-	compiler=`$(CC) --version 2>/dev/null | head -n1`; \
-	target=`$(CC) -dumpmachine 2>/dev/null || printf unknown`; \
-	sed \
-		-e "s|@LLAMA_BUILD_NUMBER@|$$bnum|" \
-		-e "s|@LLAMA_BUILD_COMMIT@|$$commit|" \
-		-e "s|@BUILD_COMPILER@|$$compiler|" \
-		-e "s|@BUILD_TARGET@|$$target|" \
-		$< > $@
-
-w64dk-build-info.cpp.o: w64dk-build-info.cpp
-
-licenses = LICENSE $(wildcard licenses/LICENSE-*)
-w64dk-license.c: $(licenses)
-	{ \
-	  printf 'const char *LICENSES[] = {\n'; \
-	  for f in $(licenses); do \
-	    printf '    "%s", (char[]){\n' "$${f##*-}"; \
-	    printf '        #embed "%s"\n' "$$f"; \
-	    printf '        , 0\n'; \
-	    printf '    },\n'; \
-	  done; \
-	  printf '    0\n};\n'; \
-	} >$@
-
-$(ui_stamp): tools/ui/package.json tools/ui/package-lock.json
-	mkdir -p $(ui_dist)
-	if test -f $(ui_dist)/index.html \
-	     && test -f $(ui_dist)/bundle.js \
-	     && test -f $(ui_dist)/bundle.css \
-	     && test -f $(ui_dist)/loading.html; then \
-	    :; \
-	  elif command -v npm >/dev/null 2>&1 \
-	       && npm --version >/dev/null 2>&1 \
-	       && (cd tools/ui && npm install && npm run build); then \
-	    :; \
-	  elif command -v wget >/dev/null 2>&1; then \
-	    ok=1; \
-	    for asset in $(ui_assets); do \
-	      wget -O $(ui_dist)/$$asset.tmp "$(ui_url)/$$asset?download=true" \
-	        && mv $(ui_dist)/$$asset.tmp $(ui_dist)/$$asset \
-	        || { ok=0; rm -f $(ui_dist)/$$asset.tmp; break; }; \
-	    done; \
-	  fi
-	test -f $(ui_dist)/index.html
-	test -f $(ui_dist)/bundle.js
-	test -f $(ui_dist)/bundle.css
-	test -f $(ui_dist)/loading.html
-	touch $@
-
-tools/ui/index.html.hpp: $(ui_stamp)
-	cd $(ui_dist) && xxd -i index.html >../../../../tools/ui/index.html.hpp
-tools/ui/bundle.js.hpp: $(ui_stamp)
-	cd $(ui_dist) && xxd -i bundle.js >../../../../tools/ui/bundle.js.hpp
-tools/ui/bundle.css.hpp: $(ui_stamp)
-	cd $(ui_dist) && xxd -i bundle.css >../../../../tools/ui/bundle.css.hpp
-tools/ui/loading.html.hpp: $(ui_stamp)
-	cd $(ui_dist) && xxd -i loading.html >../../../../tools/ui/loading.html.hpp
-tools/ui/ui.cpp.o: \
-  tools/ui/index.html.hpp \
-  tools/ui/bundle.js.hpp \
-  tools/ui/bundle.css.hpp \
-  tools/ui/loading.html.hpp
-
-llama.def: $(dll)
-	printf 'LIBRARY llama\nEXPORTS\n' >$@
-	$(CROSS)nm -j $(dll) | grep '^llama_[a-z0-9_]\+$$' | sort >>$@
-
-# Vulkan shader generation rules
-ifdef VULKAN_SDK
-
-# vulkan-shaders-gen spawns multiple glslc processes per shader, for each
-# variant, so parallelizing it spawns hundreds of glslc processes and fails
-.NOTPARALLEL: $(vk_shader_objs)
-
-$(vk_build_dir):
-	mkdir -p $@
-
-$(vk_build_dir)/%.cpp.gen.o: ggml/src/ggml-vulkan/vulkan-shaders/%.cpp | $(vk_build_dir)
-	$(HOST_CXX) -c -o $@ -std=c++17 -O2 \
-	  -Iggml/src/ggml-vulkan/vulkan-shaders $<
-$(vk_shaders_gen): $(vk_shader_gen_obj)
-	$(HOST_CXX) -o $@ $(vk_shader_gen_obj)
-
-$(vk_shader_header): $(vk_shaders_gen) $(vk_shaders)
-	mkdir -p $(vk_build_dir)/vulkan-shaders.spv
-	./$(vk_shaders_gen) \
-	  --output-dir $(vk_build_dir)/vulkan-shaders.spv \
-	  --target-hpp $(vk_shader_header)
-
-$(vk_build_dir)/%.comp.cpp: \
-  ggml/src/ggml-vulkan/vulkan-shaders/%.comp \
-  $(vk_shaders_gen) \
-  $(vk_shader_header)
-	./$(vk_shaders_gen) \
-	  --glslc $(GLSLC) \
-	  --source $< \
-	  --output-dir $(vk_build_dir)/vulkan-shaders.spv \
-	  --target-hpp $(vk_shader_header) \
-	  --target-cpp $@
-
-# Pattern rule: compile each shader .cpp to .o
-$(vk_build_dir)/%.comp.cpp.o: $(vk_build_dir)/%.comp.cpp
-	$(CROSS)$(CXX) -c -o $@ $(inc) $(def) $(CPPFLAGS) $<
-
-ggml/src/ggml-vulkan/ggml-vulkan.cpp.o: $(vk_shader_header)
-
-endif  # VULKAN_SDK