diff --git a/misc/llama-cpp/Makefile b/misc/llama-cpp/Makefile
index 3937662261bc..d4d1bf208bc8 100644
--- a/misc/llama-cpp/Makefile
+++ b/misc/llama-cpp/Makefile
@@ -1,80 +1,96 @@
 PORTNAME=	llama-cpp
 DISTVERSIONPREFIX=	b
-DISTVERSION=	9033
+DISTVERSION=	9159
 CATEGORIES=	misc # machine-learning
+MASTER_SITES=	https://huggingface.co/buckets/ggml-org/llama-ui/resolve/${DISTVERSIONPREFIX}${DISTVERSION}/:webui
+DISTFILES=	index.html:webui \
+		bundle.js:webui \
+		bundle.css:webui \
+		loading.html:webui
+DIST_SUBDIR=	${PORTNAME}
+EXTRACT_ONLY=	${DISTNAME}${EXTRACT_SUFX} \
+		nomic-ai-kompute-4565194_GH0.tar.gz
 
 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
 
 LIB_DEPENDS=	libggml-base.so:misc/ggml
 
 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 \
 		LLAMA_USE_SYSTEM_GGML
 CMAKE_OFF=	GGML_NATIVE \
 		FREEBSD_ALLOW_ADVANCED_CPU_FEATURES \
 		LLAMA_BUILD_TESTS
 CMAKE_TESTING_ON=	LLAMA_BUILD_TESTS
 
 # user for llama-server, only used when EXAMPLES=ON
 USER=		nobody
 SUB_LIST=	USER=${USER}
 
 PLIST_SUB+=	DISTVERSION=${DISTVERSION}
 
 OPTIONS_DEFINE=		CURL EXAMPLES
 OPTIONS_DEFAULT=	CURL EXAMPLES
 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
 
 BINARY_ALIAS=	git=false \
 		python=${PYTHON_CMD} # for tests
 
+post-extract:
+	@${MKDIR} ${WRKSRC}/tools/server/public
+	${CP} ${DISTDIR}/${DIST_SUBDIR}/index.html ${WRKSRC}/tools/server/public/index.html
+	${CP} ${DISTDIR}/${DIST_SUBDIR}/bundle.js ${WRKSRC}/tools/server/public/bundle.js
+	${CP} ${DISTDIR}/${DIST_SUBDIR}/bundle.css ${WRKSRC}/tools/server/public/bundle.css
+	${CP} ${DISTDIR}/${DIST_SUBDIR}/loading.html ${WRKSRC}/tools/server/public/loading.html
+
 post-patch: # set version in the code
 	@${REINPLACE_CMD} \
 		-e "s|set(BUILD_NUMBER 0)|set(BUILD_NUMBER ${DISTVERSION})|" \
 		${WRKSRC}/cmake/build-info.cmake
 
 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
 
 .include <bsd.port.options.mk>
 
 .if ${PORT_OPTIONS:MEXAMPLES}
 USE_RC_SUBR=	llama-server
 .endif
 
 # tests as of 4458: 97% tests passed, 1 tests failed out of 31, see https://github.com/ggerganov/llama.cpp/issues/11036
 
 # tests as of 4649:
 # 88% tests passed, 4 tests failed out of 32
 # The following tests FAILED:
 #         18 - test-chat (Subprocess aborted)                    main   # see https://github.com/ggerganov/llama.cpp/issues/11705
 #         24 - test-gguf (SEGFAULT)                              main
 #         25 - test-backend-ops (SEGFAULT)                       main
 #         32 - test-eval-callback (SEGFAULT)                     curl eval-callback
 
 .include <bsd.port.mk>
diff --git a/misc/llama-cpp/distinfo b/misc/llama-cpp/distinfo
index d007669d2147..f9698d42936e 100644
--- a/misc/llama-cpp/distinfo
+++ b/misc/llama-cpp/distinfo
@@ -1,5 +1,13 @@
-TIMESTAMP = 1777994381
-SHA256 (ggerganov-llama.cpp-b9033_GH0.tar.gz) = f4b2e058d0c7653e1c324d24265f6c79a214587c5d73a43b6e0366b450b00e7c
-SIZE (ggerganov-llama.cpp-b9033_GH0.tar.gz) = 34016043
-SHA256 (nomic-ai-kompute-4565194_GH0.tar.gz) = 95b52d2f0514c5201c7838348a9c3c9e60902ea3c6c9aa862193a212150b2bfc
-SIZE (nomic-ai-kompute-4565194_GH0.tar.gz) = 13540496
+TIMESTAMP = 1778834038
+SHA256 (llama-cpp/index.html) = 3ea56dac69456ecc2f31ad84d9e912155ae37f247a248d7f28107a22d6bc4af3
+SIZE (llama-cpp/index.html) = 6917
+SHA256 (llama-cpp/bundle.js) = 77f305c6b5ed882dcc49eb100b3037f1041791f72184a68e24e8f730d29c4c3a
+SIZE (llama-cpp/bundle.js) = 5274578
+SHA256 (llama-cpp/bundle.css) = 4eb1d456ea3265351c01b5aa3d7706d215d1e4cefc2fb70de58f7d703d36c17b
+SIZE (llama-cpp/bundle.css) = 505476
+SHA256 (llama-cpp/loading.html) = 2500057e39ab81518d16b28f5d019f6107b58abb47b2a30d33862d9e7b703cdc
+SIZE (llama-cpp/loading.html) = 268
+SHA256 (llama-cpp/ggerganov-llama.cpp-b9159_GH0.tar.gz) = c085b13fa80ded21fdcd9150a1617a9784466f71c6f48dbcaa6df6f9fc90d26a
+SIZE (llama-cpp/ggerganov-llama.cpp-b9159_GH0.tar.gz) = 33843673
+SHA256 (llama-cpp/nomic-ai-kompute-4565194_GH0.tar.gz) = 95b52d2f0514c5201c7838348a9c3c9e60902ea3c6c9aa862193a212150b2bfc
+SIZE (llama-cpp/nomic-ai-kompute-4565194_GH0.tar.gz) = 13540496