diff --git a/misc/llama-cpp/Makefile b/misc/llama-cpp/Makefile
index 1eeab118f76e..ed8f01264362 100644
--- a/misc/llama-cpp/Makefile
+++ b/misc/llama-cpp/Makefile
@@ -1,55 +1,55 @@
 PORTNAME=	llama-cpp
 DISTVERSIONPREFIX=	b
-DISTVERSION=	4409
+DISTVERSION=	4418
 CATEGORIES=	misc # machine-learning
 
 MAINTAINER=	yuri@FreeBSD.org
 COMMENT=	Facebook's LLaMA model in C/C++ # '
 WWW=		https://github.com/ggerganov/llama.cpp
 
 LICENSE=	MIT
 LICENSE_FILE=	${WRKSRC}/LICENSE
 
 BROKEN_armv7=	clang crashes, see https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=278810
 BROKEN_i386=	compilation fails, see https://github.com/ggerganov/llama.cpp/issues/9545
 
 USES=		cmake:testing compiler:c++11-lang python:run shebangfix
 USE_LDCONFIG=	yes
 
 USE_GITHUB=	yes
 GH_ACCOUNT=	ggerganov
 GH_PROJECT=	llama.cpp
 GH_TUPLE=	nomic-ai:kompute:4565194:kompute/kompute
 
 SHEBANG_GLOB=	*.py
 
 CMAKE_ON=	BUILD_SHARED_LIBS
 CMAKE_OFF=	LLAMA_BUILD_TESTS
 CMAKE_TESTING_ON=	LLAMA_BUILD_TESTS
 
 OPTIONS_DEFINE=		CURL EXAMPLES VULKAN
 OPTIONS_DEFAULT=	CURL VULKAN
 OPTIONS_SUB=		yes
 
 CURL_DESCR=		Use libcurl to download model from an URL
 CURL_CMAKE_BOOL=	LLAMA_CURL
 CURL_USES=		localbase
 CURL_LIB_DEPENDS=	libcurl.so:ftp/curl
 
 EXAMPLES_CMAKE_BOOL=	LLAMA_BUILD_EXAMPLES
 
 VULKAN_DESC=		Vulkan GPU offload support
 VULKAN_CMAKE_BOOL=	GGML_VULKAN
 VULKAN_BUILD_DEPENDS=	glslc:graphics/shaderc \
 			vulkan-headers>0:graphics/vulkan-headers
 VULKAN_LIB_DEPENDS=	libvulkan.so:graphics/vulkan-loader
 
 BINARY_ALIAS=	git=false
 
 do-test-ci: # build of tests fails, see https://github.com/ggerganov/llama.cpp/issues/10955
 	@cd ${WRKSRC} && \
 		${SETENV} ${MAKE_ENV} bash ci/run.sh ./tmp/results ./tmp/mnt
 
 # tests as of 4404: 97% tests passed, 1 tests failed out of 31, see https://github.com/ggerganov/llama.cpp/issues/11036
 
 .include <bsd.port.mk>
diff --git a/misc/llama-cpp/distinfo b/misc/llama-cpp/distinfo
index 1077df5ceb17..45282e0c4897 100644
--- a/misc/llama-cpp/distinfo
+++ b/misc/llama-cpp/distinfo
@@ -1,5 +1,5 @@
-TIMESTAMP = 1735962918
-SHA256 (ggerganov-llama.cpp-b4409_GH0.tar.gz) = 71a62315c73b1f00bdd5fdec7d16a77a0d6e1150171953fde230f04e7d5adfa1
-SIZE (ggerganov-llama.cpp-b4409_GH0.tar.gz) = 20608272
+TIMESTAMP = 1736116712
+SHA256 (ggerganov-llama.cpp-b4418_GH0.tar.gz) = b1d13215cbdb076ead3ed86043edc5c96999a204b95ad778fce8cd229d4d5427
+SIZE (ggerganov-llama.cpp-b4418_GH0.tar.gz) = 20608757
 SHA256 (nomic-ai-kompute-4565194_GH0.tar.gz) = 95b52d2f0514c5201c7838348a9c3c9e60902ea3c6c9aa862193a212150b2bfc
 SIZE (nomic-ai-kompute-4565194_GH0.tar.gz) = 13540496
diff --git a/misc/llama-cpp/files/patch-ggml_src_CMakeLists.txt b/misc/llama-cpp/files/patch-ggml_src_CMakeLists.txt
new file mode 100644
index 000000000000..8135824e1b8b
--- /dev/null
+++ b/misc/llama-cpp/files/patch-ggml_src_CMakeLists.txt
@@ -0,0 +1,20 @@
+- workaround for https://github.com/ggerganov/llama.cpp/issues/11095
+
+--- ggml/src/CMakeLists.txt.orig	2025-01-06 00:37:35 UTC
++++ ggml/src/CMakeLists.txt
+@@ -152,15 +152,6 @@ endif()
+ # posix_memalign came in POSIX.1-2001 / SUSv3
+ # M_PI is an XSI extension since POSIX.1-2001 / SUSv3, came in XPG1 (1985)
+ 
+-# Somehow in OpenBSD whenever POSIX conformance is specified
+-# some string functions rely on locale_t availability,
+-# which was introduced in POSIX.1-2008, forcing us to go higher
+-if (CMAKE_SYSTEM_NAME MATCHES "OpenBSD")
+-    add_compile_definitions(_XOPEN_SOURCE=700)
+-else()
+-    add_compile_definitions(_XOPEN_SOURCE=600)
+-endif()
+-
+ # Data types, macros and functions related to controlling CPU affinity and
+ # some memory allocation are available on Linux through GNU extensions in libc
+ if (CMAKE_SYSTEM_NAME MATCHES "Linux" OR CMAKE_SYSTEM_NAME MATCHES "Android")
diff --git a/misc/llama-cpp/pkg-message b/misc/llama-cpp/pkg-message
new file mode 100644
index 000000000000..071e82665d9a
--- /dev/null
+++ b/misc/llama-cpp/pkg-message
@@ -0,0 +1,17 @@
+[
+{ type: install
+  message: <<EOM
+You installed LLaMA-cpp: Facebook's LLaMA model runner.
+
+In order to experience LLaMA-cpp please download some
+AI model in the GGUF format, for example from huggingface.com,
+run the script below, and open localhost:9011 in your browser
+to communicate with this AI model.
+
+$ llama-server -m $MODEL \
+  --host 0.0.0.0 \
+  --port 9011
+
+EOM
+}
+]