diff --git a/misc/py-llama-cpp-python/Makefile b/misc/py-llama-cpp-python/Makefile
index 2cc013916c94..6b4387d5bf18 100644
--- a/misc/py-llama-cpp-python/Makefile
+++ b/misc/py-llama-cpp-python/Makefile
@@ -1,41 +1,41 @@
 PORTNAME=	llama-cpp-python
 DISTVERSIONPREFIX=	v
-DISTVERSION=	0.3.2
+DISTVERSION=	0.3.16
 CATEGORIES=	misc # machine-learning
 PKGNAMEPREFIX=	${PYTHON_PKGNAMEPREFIX}
 
 MAINTAINER=	yuri@FreeBSD.org
 COMMENT=	Python bindings for the llama.cpp library
 WWW=		https://llama-cpp-python.readthedocs.io/en/latest/ \
 		https://github.com/abetlen/llama-cpp-python
 
 LICENSE=	MIT
 LICENSE_FILE=	${WRKSRC}/LICENSE.md
 
-BUILD_DEPENDS=	${PYTHON_PKGNAMEPREFIX}scikit-build-core>0:devel/py-scikit-build-core@${PY_FLAVOR} \
+BUILD_DEPENDS=	${PYTHON_PKGNAMEPREFIX}scikit-build-core>=0.9.2:devel/py-scikit-build-core@${PY_FLAVOR} \
 		cmake:devel/cmake-core
 LIB_DEPENDS=	libvulkan.so:graphics/vulkan-loader
 RUN_DEPENDS=	${PYTHON_PKGNAMEPREFIX}diskcache>=5.6.1:devel/py-diskcache@${PY_FLAVOR} \
 		${PYTHON_PKGNAMEPREFIX}Jinja2>=2.11.3:devel/py-Jinja2@${PY_FLAVOR} \
 		${PYNUMPY} \
 		${PYTHON_PKGNAMEPREFIX}typing-extensions>=4.5.0:devel/py-typing-extensions@${PY_FLAVOR}
 RUN_DEPENDS+=	${PYTHON_PKGNAMEPREFIX}fastapi>=0.100.0:www/py-fastapi@${PY_FLAVOR} \
 		${PYTHON_PKGNAMEPREFIX}pydantic-settings>=2.0.1:devel/py-pydantic-settings@${PY_FLAVOR} \
 		${PYTHON_PKGNAMEPREFIX}sse-starlette>=1.6.1:www/py-sse-starlette@${PY_FLAVOR} \
 		${PYTHON_PKGNAMEPREFIX}starlette-context>=0.3.6:www/py-starlette-context@${PY_FLAVOR} \
 		${PYTHON_PKGNAMEPREFIX}uvicorn>=0.22.0:www/py-uvicorn@${PY_FLAVOR} \
 		${PYTHON_PKGNAMEPREFIX}pyyaml>=5.1:devel/py-pyyaml@${PY_FLAVOR}
 TEST_DEPENDS=	${PYTHON_PKGNAMEPREFIX}httpx>=0.24.1:www/py-httpx@${PY_FLAVOR} \
 		${PYTHON_PKGNAMEPREFIX}huggingface-hub>=0.23.0:misc/py-huggingface-hub@${PY_FLAVOR} \
 		${PYTHON_PKGNAMEPREFIX}scipy>=1.10:science/py-scipy@${PY_FLAVOR}
 
 USES=		python shebangfix
 USE_PYTHON=	pep517 autoplist pytest
 
 USE_GITHUB=	yes
 GH_ACCOUNT=	abetlen
-GH_TUPLE=	ggerganov:llama.cpp:74d73dc:cpp/vendor/llama.cpp
+GH_TUPLE=	ggerganov:llama.cpp:b6598:cpp/vendor/llama.cpp
 
 SHEBANG_GLOB=	*.py
 
 .include <bsd.port.mk>
diff --git a/misc/py-llama-cpp-python/distinfo b/misc/py-llama-cpp-python/distinfo
index a25d0e5c74ba..7f4849c553a3 100644
--- a/misc/py-llama-cpp-python/distinfo
+++ b/misc/py-llama-cpp-python/distinfo
@@ -1,5 +1,5 @@
-TIMESTAMP = 1731907964
-SHA256 (abetlen-llama-cpp-python-v0.3.2_GH0.tar.gz) = 491057624a095ff6d4eae1bcc139b579f72d03d9deee1a2ab05d2f980ec824ed
-SIZE (abetlen-llama-cpp-python-v0.3.2_GH0.tar.gz) = 274530
-SHA256 (ggerganov-llama.cpp-74d73dc_GH0.tar.gz) = 9cc0aad9746b93b461c53d7038211ea7cbe70df2aa1a23b2daac07af935f6990
-SIZE (ggerganov-llama.cpp-74d73dc_GH0.tar.gz) = 19564318
+TIMESTAMP = 1758915707
+SHA256 (abetlen-llama-cpp-python-v0.3.16_GH0.tar.gz) = d6ae5a6ac40dda4d14c6bb8f5e9504d28f442cf810263661d457c948c386f2a4
+SIZE (abetlen-llama-cpp-python-v0.3.16_GH0.tar.gz) = 279565
+SHA256 (ggerganov-llama.cpp-b6598_GH0.tar.gz) = cd296792f49695bd44e885a5c1fcefe4ef72f3da7f8933be1378f944116515a3
+SIZE (ggerganov-llama.cpp-b6598_GH0.tar.gz) = 25833773
diff --git a/misc/py-llama-cpp-python/files/patch-vendor_llama.cpp_ggml_src_ggml-cpu_ggml-cpu-impl.h b/misc/py-llama-cpp-python/files/patch-vendor_llama.cpp_ggml_src_ggml-cpu_ggml-cpu-impl.h
deleted file mode 100644
index 376142c64242..000000000000
--- a/misc/py-llama-cpp-python/files/patch-vendor_llama.cpp_ggml_src_ggml-cpu_ggml-cpu-impl.h
+++ /dev/null
@@ -1,11 +0,0 @@
---- vendor/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-impl.h.orig	2025-08-02 23:17:06 UTC
-+++ vendor/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-impl.h
-@@ -319,8 +319,6 @@ inline static int32x4_t ggml_vdotq_s32(int32x4_t acc, 
- #else
- #ifdef __POWER9_VECTOR__
- #include <altivec.h>
--#undef bool
--#define bool _Bool
- #else
- #if defined(_MSC_VER) || defined(__MINGW32__)
- #include <intrin.h>