diff --git a/sci-misc/llama-cpp/Manifest b/sci-misc/llama-cpp/Manifest
index a7549fee0a..79e6274de3 100644
--- a/sci-misc/llama-cpp/Manifest
+++ b/sci-misc/llama-cpp/Manifest
@@ -1,7 +1,5 @@
 DIST ggml-org_models_tinyllamas_stories15M-q4_0-99dd1a73db5a37100bd4ae633f4cfce6560e1567.gguf 19077344 BLAKE2B 16e65adf9785e3091c51f1de59e5580f93fb47f79961513aeb3dbb8a0f5930f7120f0304f0f293a006170805e2b70ee1fcff0496b63356323d32c2caa55be8a8 SHA512 f9944886089958e0d97b1906cfd45020e0821c65429346e76fae29136c634ae5d039dffbae5933a95b0674f4acd87b656feb9f9e1b16dd434c5c9b5886f4f617
 DIST llama-b10809-ui.tar.gz 3084118 BLAKE2B 3cd4e0bd708fb673507b1480e7e814bcb25b84ca096c22d491fbb02ea28523f28517ad0cbf15a7c689801599206963b1268239eba6d028a9343956710c39d07c SHA512 485b4331f547bcf53fa40691e43ca97de90301f656b8176c2ae5091bb2c6d0dcee57439f12a02a06fa5a397f5aba2481e8614c2a06da1396fa8efa58b704c973
-DIST llama-b10948-ui.tar.gz 3086256 BLAKE2B 537e559af6fa3e09c4b630e1fbad6e79abc172da45abf369ae477133f6d17049dc63cfa79ebd7fd711e46ffcc7ace61127e6fd013564928ab03c46ff06267e91 SHA512 f109da63d3f6d2e0f9f6c1d5a6f6b41535f2c2a2a52fa1ed0f690c5bb34778a898ceefe3d09c7805e9476fcda410d5fd372f77bdb5c19b00e1ef38ae5b1830f0
 DIST llama-b10964-ui.tar.gz 3084524 BLAKE2B b53a3fdcb8a0dd132c125409ec145b927493d929a1a4e1f266f2f6ea08aae684dbaecadf039b7f6bab05aa93890ee9e5a5e926ccc1f0f32d7b1a74e1ba77c1b7 SHA512 3524c4ff43bf334cc1ee75e03bf5a33dba884a20489ac8ee6bd5d5d27060bc183f8939d301295576c3dbf9263dd9a55022c3e683e2e88b2686c1fe81311d4d02
 DIST llama-cpp-0.4.0.tar.gz 37288249 BLAKE2B a0cd0c8251b4dc32681426e323e52b03c9a733a3d29d8e733334702c2082b2aff664906811dee0298ea950a28fec40771e5c27aa5a2c32b7f7b9b9cab512249d SHA512 12592516c0ac70ed53d64fc0c9f82bbb156defa507c7364f1a753eb29f3498e0e13a282e139c2eff7d2eb00921b4ef107b2fa90a5b66f8a3c45409d4fcb0834f
-DIST llama-cpp-0.4.0_p10948.tar.gz 37412172 BLAKE2B 814bef31c941a9d04a160ca76b701bbc4038938457f261eabe57ba9e27a20d112365fd11f4718b7c8f5be85d9a3f682dd750a1f987e8c0f95dd8ba7f770c86d0 SHA512 4dda47454faa68826949def46e36226c40a9dafe69ce9edfbc6b9c9a1c4b4b54c7531adc9723c50ebfd1937d523a33db97754c753003cd4b21c31cd89c85f50f
 DIST llama-cpp-0.4.1.tar.gz 37422659 BLAKE2B beefc257a01a11c8ecac25d963d37b05679715bece94d2f64c391c6158c7f81179873b66c36661159ae0fc126c89b4df6d47e8fe853b7e052d2f1297de13155f SHA512 b82f755c40897d273ec3784f4f22b02f6a5f5f7b3c2a4240495d13a4617575c9aa69597de6a4f0b9a6568f51cc976e0b72ac847d83c2f953ed7fa52365e71d9a
diff --git a/sci-misc/llama-cpp/llama-cpp-0.4.0_p10948.ebuild b/sci-misc/llama-cpp/llama-cpp-0.4.0_p10948.ebuild
deleted file mode 100644
index 16f3579b11..0000000000
--- a/sci-misc/llama-cpp/llama-cpp-0.4.0_p10948.ebuild
+++ /dev/null
@@ -1,237 +0,0 @@
-# Copyright 2026 Gentoo Authors
-# Distributed under the terms of the GNU General Public License v2
-
-EAPI=8
-
-ROCM_VERSION="7.1"
-
-inherit cmake cuda rocm linux-info toolchain-funcs
-
-TINY_LLAMAS_COMMIT="99dd1a73db5a37100bd4ae633f4cfce6560e1567"
-
-DESCRIPTION="Port of Facebook's LLaMA model in C/C++"
-HOMEPAGE="https://github.com/ggml-org/llama.cpp"
-
-if [[ ${PV} == *9999* ]]; then
-	inherit git-r3
-	EGIT_REPO_URI="https://github.com/ggml-org/llama.cpp.git"
-	LLAMA_BUILD_IS_DEV=ON
-	LLAMA_BUILD_NUMBER="b0000"
-	LLAMA_UI_VERSION="b10948"
-	MY_PV="${PV}"
-else
-	if [[ $(ver_cut 4) == "p" ]]; then
-		LLAMA_BUILD_IS_DEV=ON
-		LLAMA_BUILD_NUMBER="b$(ver_cut 5)"
-		MY_PV="${LLAMA_BUILD_NUMBER}"
-		S="${WORKDIR}/llama.cpp-${MY_PV}"
-	else
-		LLAMA_BUILD_IS_DEV=OFF
-		# Set from https://github.com/ggml-org/llama.cpp/releases/download/v${PV}/nightly-tag.txt
-		LLAMA_BUILD_NUMBER="b10948"
-		MY_PV="v${PV}"
-		S="${WORKDIR}/llama.cpp-${PV}"
-	fi
-	LLAMA_UI_VERSION="${LLAMA_BUILD_NUMBER}"
-	SRC_URI="https://github.com/ggml-org/llama.cpp/archive/refs/tags/${MY_PV}.tar.gz -> ${P}.tar.gz"
-	KEYWORDS="~amd64"
-fi
-
-SRC_URI+="
-	https://github.com/ggml-org/llama.cpp/releases/download/${LLAMA_UI_VERSION}/llama-${LLAMA_UI_VERSION}-ui.tar.gz
-	examples? (
-		https://huggingface.co/ggml-org/tiny-llamas/resolve/${TINY_LLAMAS_COMMIT}/stories15M-q4_0.gguf
-			-> ggml-org_models_tinyllamas_stories15M-q4_0-${TINY_LLAMAS_COMMIT}.gguf
-	)
-"
-
-LICENSE="MIT"
-SLOT="0"
-X86_CPU_FLAGS=(
-	amx_bf16
-	amx_int8
-	amx_tile
-	avx
-	avx2
-	avx512_bf16
-	avx512_vnni
-	avx512f
-	avx512vbmi
-	avx_vnni
-	bmi2
-	f16c
-	fma3
-	sse4_2
-)
-CPU_FLAGS=( "${X86_CPU_FLAGS[@]/#/cpu_flags_x86_}" )
-
-# wwma USE explained here: https://github.com/ggml-org/llama.cpp/blob/master/docs/build.md#hip
-IUSE="${CPU_FLAGS[*]} blis cuda curl examples flexiblas openblas opencl +openmp openssl rocm vulkan wmma"
-
-REQUIRED_USE="
-	?? (
-		openblas
-		blis
-		flexiblas
-	)
-	rocm? ( ${ROCM_REQUIRED_USE} )
-	wmma? (
-		rocm
-	)
-"
-
-# curl is needed for pulling models from huggingface
-# numpy is used by convert_hf_to_gguf.py
-CDEPEND="
-	curl? ( net-misc/curl:= )
-	openblas? ( sci-libs/openblas:= )
-	openmp? ( llvm-runtimes/openmp:= )
-	blis? ( sci-libs/blis:= )
-	flexiblas? ( sci-libs/flexiblas:= )
-	rocm? (
-		>=dev-util/hip-${ROCM_VERSION}:=
-		>=sci-libs/hipBLAS-${ROCM_VERSION}:=[${ROCM_USEDEP}]
-		wmma? (
-			>=sci-libs/rocWMMA-${ROCM_VERSION}:=[${ROCM_USEDEP}]
-		)
-	)
-	cuda? ( dev-util/nvidia-cuda-toolkit:= )
-	openssl? ( dev-libs/openssl:= )
-"
-DEPEND="${CDEPEND}
-	opencl? ( dev-util/opencl-headers )
-	vulkan? (
-		dev-util/spirv-headers
-		dev-util/vulkan-headers
-	)
-"
-RDEPEND="${CDEPEND}
-	dev-python/numpy
-	opencl? ( dev-libs/opencl-icd-loader )
-	vulkan? ( media-libs/vulkan-loader )
-"
-BDEPEND="media-libs/shaderc"
-
-pkg_pretend() {
-	[[ ${MERGE_TYPE} != binary ]] && use openmp && tc-check-openmp
-}
-
-pkg_setup() {
-	if use rocm; then
-		linux-info_pkg_setup
-		if linux-info_get_any_version && linux_config_exists; then
-			if ! linux_chkconfig_present HSA_AMD_SVM; then
-				ewarn "To use ROCm/HIP, you need to have HSA_AMD_SVM option enabled in your kernel."
-			fi
-		fi
-	fi
-
-	[[ ${MERGE_TYPE} != binary ]] && use openmp && tc-check-openmp
-}
-
-src_unpack() {
-	[[ ${PV} == *9999* ]] && git-r3_src_unpack
-	default
-}
-
-src_prepare() {
-	use cuda && cuda_src_prepare
-	cmake_src_prepare
-	if use examples; then
-		mkdir -p "${BUILD_DIR}/tinyllamas" || die
-		cp "${DISTDIR}/ggml-org_models_tinyllamas_stories15M-q4_0-${TINY_LLAMAS_COMMIT}.gguf" \
-			"${BUILD_DIR}/tinyllamas/stories15M-q4_0.gguf" || die
-	fi
-
-	# Move web UI assets where they belong, bug #979245
-	mkdir -p "${S}"/tools/ui || die
-	cp -a "${WORKDIR}"/llama-${LLAMA_UI_VERSION} "${S}"/tools/ui/dist || die
-}
-
-src_configure() {
-	local mycmakeargs=(
-		-DCMAKE_BUILD_WITH_INSTALL_RPATH=ON
-		-DLLAMA_BUILD_IS_DEV=${LLAMA_BUILD_IS_DEV}
-		-DLLAMA_BUILD_TESTS=OFF
-		-DLLAMA_BUILD_EXAMPLES=$(usex examples)
-		-DLLAMA_BUILD_SERVER=ON
-		-DBUILD_NUMBER="${LLAMA_BUILD_NUMBER}"
-		-DLLAMA_CURL=$(usex curl)
-		-DLLAMA_OPENSSL=$(usex openssl)
-
-		# avoid clashing with whisper.cpp
-		-DCMAKE_INSTALL_LIBDIR="${EPREFIX}/usr/$(get_libdir)/llama.cpp"
-		-DCMAKE_INSTALL_RPATH="${EPREFIX}/usr/$(get_libdir)/llama.cpp"
-	)
-
-	# GGML backends
-	mycmakeargs+=(
-		-DGGML_NATIVE=OFF	# don't set march
-
-		# CPU Flags
-		-DGGML_SSE42=$(usex cpu_flags_x86_sse4_2)
-		-DGGML_AVX=$(usex cpu_flags_x86_avx)
-		-DGGML_AVX_VNNI=$(usex cpu_flags_x86_avx_vnni)
-		-DGGML_AVX2=$(usex cpu_flags_x86_avx2)
-		-DGGML_BMI2=$(usex cpu_flags_x86_bmi2)
-		-DGGML_AVX512=$(usex cpu_flags_x86_avx512f)
-		-DGGML_AVX512_VBMI=$(usex cpu_flags_x86_avx512vbmi)
-		-DGGML_AVX512_VNNI=$(usex cpu_flags_x86_avx512_vnni)
-		-DGGML_AVX512_BF16=$(usex cpu_flags_x86_avx512_bf16)
-		-DGGML_FMA=$(usex cpu_flags_x86_fma3)
-		-DGGML_F16C=$(usex cpu_flags_x86_f16c)
-		-DGGML_AMX_TILE=$(usex cpu_flags_x86_amx_tile)
-		-DGGML_AMX_INT8=$(usex cpu_flags_x86_amx_int8)
-		-DGGML_AMX_BF16=$(usex cpu_flags_x86_amx_bf16)
-
-		-DGGML_CUDA=$(usex cuda)
-		-DGGML_VULKAN=$(usex vulkan)
-		-DGGML_OPENMP=$(usex openmp)
-		-DGGML_RPC=ON
-		-DGGML_OPENCL=$(usex opencl)
-	)
-
-	if use openblas ; then
-		mycmakeargs+=(
-			-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS
-		)
-	fi
-
-	if use blis ; then
-		mycmakeargs+=(
-			-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=FLAME
-		)
-	fi
-
-	if use flexiblas; then
-		mycmakeargs+=(
-			-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=FlexiBLAS
-		)
-	fi
-
-	if use cuda; then
-		local -x CUDAHOSTCXX="$(cuda_gccdir)"
-		# tries to recreate dev symlinks
-		cuda_add_sandbox
-		addpredict "/dev/char/"
-	fi
-
-	if use rocm; then
-		rocm_use_hipcc
-		mycmakeargs+=(
-			-DGGML_HIP=ON
-			-DGPU_TARGETS=$(get_amdgpu_flags)
-			-DGGML_HIP_ROCWMMA_FATTN=$(usex wmma)
-		)
-	fi
-
-	cmake_src_configure
-}
-
-src_install() {
-	cmake_src_install
-	dobin "${BUILD_DIR}/bin/ggml-rpc-server"
-
-	# avoid clashing with whisper.cpp
-	rm -r "${ED}/usr/include" || die
-}