diff --git a/.docker/rocmfp4-compile.sh b/.docker/rocmfp4-compile.sh
new file mode 100755
index 000000000000..7547ab5f0960
--- /dev/null
+++ b/.docker/rocmfp4-compile.sh
@@ -0,0 +1,45 @@
+#!/usr/bin/env bash
+# Shared compile logic for backend/Dockerfile.rocmfp4.
+# Sourced (via bind mount) from both builder-fromsource and builder-prebuilt stages.
+
+set -euxo pipefail
+
+# Docker ARG defaults arrive as empty strings, and an empty-but-defined variable
+# still beats Make's ?= assignment. Drop them so the Makefile's pin wins unless a
+# local build genuinely overrides it.
+[ -n "${LLAMA_REPO:-}" ] || unset LLAMA_REPO || true
+[ -n "${ROCMFP4_VERSION:-}" ] || unset ROCMFP4_VERSION || true
+
+export CCACHE_DIR=/root/.ccache
+ccache --max-size=5G || true
+ccache -z || true
+
+export CMAKE_ARGS="${CMAKE_ARGS:-} -DCMAKE_C_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_COMPILER_LAUNCHER=ccache -DCMAKE_CUDA_COMPILER_LAUNCHER=ccache"
+
+if [[ -n "${CUDA_DOCKER_ARCH:-}" ]]; then
+  CUDA_ARCH_ESC="${CUDA_DOCKER_ARCH//;/\\;}"
+  export CMAKE_ARGS="${CMAKE_ARGS} -DCMAKE_CUDA_ARCHITECTURES=${CUDA_ARCH_ESC}"
+  echo "CMAKE_ARGS(env) = ${CMAKE_ARGS}"
+  rm -rf /LocalAI/backend/cpp/rocmfp4-*-build
+fi
+
+cd /LocalAI/backend/cpp/rocmfp4
+
+if [ -z "${BUILD_TYPE:-}" ]; then
+  # Pure CPU image: one ggml CPU_ALL_VARIANTS build replaces the per-microarch binaries.
+  # arm64: the armv9.2 SME variants need gcc-14 (gcc-13 rejects +sme).
+  if [ "${TARGETARCH}" = "arm64" ]; then
+    apt-get update -qq && apt-get install -y -qq gcc-14 g++-14
+    export CC=gcc-14 CXX=g++-14
+  fi
+  make rocmfp4-cpu-all
+else
+  # GPU build (cublas/hipblas/sycl/vulkan/...): single fallback CPU build, the accelerator
+  # does the compute. Keeps the GPU compile from also building the CPU variant matrix and
+  # avoids the gcc-14 apt step on GPU base images such as nvidia l4t.
+  make rocmfp4-fallback
+fi
+make rocmfp4-grpc
+make rocmfp4-rpc-server
+
+ccache -s || true
diff --git a/.github/backend-matrix.yml b/.github/backend-matrix.yml
index e7bada4b4e91..0052634a69d6 100644
--- a/.github/backend-matrix.yml
+++ b/.github/backend-matrix.yml
@@ -2243,6 +2243,20 @@ include:
     dockerfile: "./backend/Dockerfile.llama-cpp"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: 'hipblas'
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-gpu-rocm-hipblas-rocmfp4'
+    builder-base-image: 'quay.io/go-skynet/ci-cache:base-grpc-rocm-amd64'
+    runs-on: 'ubuntu-latest'
+    base-image: "rocm/dev-ubuntu-24.04:7.2.1"
+    skip-drivers: 'false'
+    backend: "rocmfp4"
+    dockerfile: "./backend/Dockerfile.rocmfp4"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: 'hipblas'
     cuda-major-version: ""
     cuda-minor-version: ""
diff --git a/.github/workflows/bump_deps.yaml b/.github/workflows/bump_deps.yaml
index 069b094a04d4..b9187732d7ae 100644
--- a/.github/workflows/bump_deps.yaml
+++ b/.github/workflows/bump_deps.yaml
@@ -26,6 +26,10 @@ jobs:
             variable: "BONSAI_VERSION"
             branch: "prism"
             file: "backend/cpp/bonsai/Makefile"
+          - repository: "walcz-de/llama.cpp-ROCmFP4"
+            variable: "ROCMFP4_VERSION"
+            branch: "rocmfp4"
+            file: "backend/cpp/rocmfp4/Makefile"
           - repository: "antirez/ds4"
             variable: "DS4_VERSION"
             branch: "main"
diff --git a/Makefile b/Makefile
index c68cd55223e0..ddde4eb25ab2 100644
--- a/Makefile
+++ b/Makefile
@@ -1,5 +1,5 @@
 # Disable parallel execution for backend builds
-.NOTPARALLEL: backends/diffusers backends/llama-cpp backends/turboquant backends/bonsai backends/outetts backends/piper backends/stablediffusion-ggml backends/trellis2cpp backends/trellis2cpp-darwin backends/whisper backends/crispasr backends/parakeet-cpp backends/moss-transcribe-cpp backends/nemo-speech-cpp backends/faster-whisper backends/silero-vad backends/local-store backends/valkey-store backends/cloud-proxy backends/huggingface backends/rfdetr backends/rfdetr-cpp backends/insightface backends/speaker-recognition backends/kitten-tts backends/kokoro backends/chatterbox backends/llama-cpp-darwin backends/neutts build-darwin-python-backend build-darwin-go-backend backends/mlx backends/diffuser-darwin backends/mlx-vlm backends/mlx-audio backends/mlx-distributed backends/stablediffusion-ggml-darwin backends/vllm backends/vllm-omni backends/longcat-video backends/sglang backends/moonshine backends/pocket-tts backends/qwen-tts backends/faster-qwen3-tts backends/qwen-asr backends/nemo backends/voxcpm backends/whisperx backends/ace-step backends/acestep-cpp backends/fish-speech backends/voxtral backends/opus backends/trl backends/llama-cpp-quantization backends/kokoros backends/sam3-cpp backends/qwen3-tts-cpp backends/moss-tts-cpp backends/magpie-tts-cpp backends/vllm-cpp backends/omnivoice-cpp backends/vibevoice-cpp backends/localvqe backends/tinygrad backends/sherpa-onnx backends/ds4 backends/ds4-darwin backends/liquid-audio backends/supertonic backends/depth-anything-cpp backends/privacy-filter backends/privacy-filter-darwin backends/audio-cpp backends/audio-cpp-darwin
+.NOTPARALLEL: backends/rocmfp4 backends/diffusers backends/llama-cpp backends/turboquant backends/bonsai backends/outetts backends/piper backends/stablediffusion-ggml backends/trellis2cpp backends/trellis2cpp-darwin backends/whisper backends/crispasr backends/parakeet-cpp backends/moss-transcribe-cpp backends/nemo-speech-cpp backends/faster-whisper backends/silero-vad backends/local-store backends/valkey-store backends/cloud-proxy backends/huggingface backends/rfdetr backends/rfdetr-cpp backends/insightface backends/speaker-recognition backends/kitten-tts backends/kokoro backends/chatterbox backends/llama-cpp-darwin backends/neutts build-darwin-python-backend build-darwin-go-backend backends/mlx backends/diffuser-darwin backends/mlx-vlm backends/mlx-audio backends/mlx-distributed backends/stablediffusion-ggml-darwin backends/vllm backends/vllm-omni backends/longcat-video backends/sglang backends/moonshine backends/pocket-tts backends/qwen-tts backends/faster-qwen3-tts backends/qwen-asr backends/nemo backends/voxcpm backends/whisperx backends/ace-step backends/acestep-cpp backends/fish-speech backends/voxtral backends/opus backends/trl backends/llama-cpp-quantization backends/kokoros backends/sam3-cpp backends/qwen3-tts-cpp backends/moss-tts-cpp backends/magpie-tts-cpp backends/vllm-cpp backends/omnivoice-cpp backends/vibevoice-cpp backends/localvqe backends/tinygrad backends/sherpa-onnx backends/ds4 backends/ds4-darwin backends/liquid-audio backends/supertonic backends/depth-anything-cpp backends/privacy-filter backends/privacy-filter-darwin backends/audio-cpp backends/audio-cpp-darwin
 
 GOCMD=go
 GOTEST=$(GOCMD) test
@@ -750,6 +750,16 @@ test-extra-backend-turboquant: docker-build-turboquant
 	BACKEND_TEST_CACHE_TYPE_V=turbo3 \
 	$(MAKE) test-extra-backend
 
+## rocmfp4: exercises the llama.cpp-fork backend with a real ROCmFP4 model — the
+## published Strix Halo build of Qwen3.8-27B, whose weight quants (ggml types
+## 100/103) are *only* decodable by this fork. Loading it is what makes the
+## backend distinct from stock llama-cpp. Note the artifact is 13.75 GiB; this
+## target is meant for a workstation with the weights cached, not for slim CI.
+test-extra-backend-rocmfp4: docker-build-rocmfp4
+	BACKEND_IMAGE=local-ai-backend:rocmfp4 \
+	BACKEND_TEST_MODEL_URL=https://huggingface.co/kingjones777/Qwen3.8-27B-ROCmFP4-STRIX-MTP-GGUF/resolve/main/Qwen3.8-27B-Q4_0_ROCMFP4_STRIX.gguf \
+	$(MAKE) test-extra-backend
+
 ## bonsai: exercises the llama.cpp-fork backend with a real Q1_0 (1-bit) model —
 ## the PrismML Bonsai-8B GGUF, whose weight quant is *only* decodable by the fork's
 ## Q1_0 kernels. Loading it is what makes this backend distinct from stock llama-cpp;
@@ -1284,6 +1294,9 @@ BACKEND_TURBOQUANT = turboquant|turboquant|.|false|false
 # weight-quant kernels the Bonsai / Ternary-Bonsai models ship in. Reuses
 # backend/cpp/llama-cpp grpc-server sources via a thin wrapper Makefile.
 BACKEND_BONSAI = bonsai|bonsai|.|false|false
+# rocmfp4 is a llama.cpp fork carrying the ROCmFP4 / ROCmFPx weight formats
+# (ggml types 100-107) for AMD RDNA3.5 APUs. Stock llama.cpp rejects those types.
+BACKEND_ROCMFP4 = rocmfp4|rocmfp4|.|false|false
 # ds4 is antirez/ds4, a DeepSeek V4 Flash-specific inference engine.
 # Single-model; hardware-only validation lives at tests/e2e-backends/
 # (BACKEND_BINARY mode); see docs/superpowers/plans/2026-05-11-ds4-backend.md.
@@ -1399,6 +1412,7 @@ $(eval $(call generate-docker-build-target,$(BACKEND_LLAMA_CPP)))
 $(eval $(call generate-docker-build-target,$(BACKEND_IK_LLAMA_CPP)))
 $(eval $(call generate-docker-build-target,$(BACKEND_TURBOQUANT)))
 $(eval $(call generate-docker-build-target,$(BACKEND_BONSAI)))
+$(eval $(call generate-docker-build-target,$(BACKEND_ROCMFP4)))
 $(eval $(call generate-docker-build-target,$(BACKEND_DS4)))
 $(eval $(call generate-docker-build-target,$(BACKEND_PRIVACY_FILTER)))
 $(eval $(call generate-docker-build-target,$(BACKEND_AUDIO_CPP)))
@@ -1471,7 +1485,7 @@ $(eval $(call generate-docker-build-target,$(BACKEND_SUPERTONIC)))
 docker-save-%: backend-images
 	docker save local-ai-backend:$* -o backend-images/$*.tar
 
-docker-build-backends: docker-build-llama-cpp docker-build-ik-llama-cpp docker-build-turboquant docker-build-bonsai docker-build-ds4 docker-build-rerankers docker-build-vllm docker-build-vllm-omni docker-build-longcat-video docker-build-sglang docker-build-transformers docker-build-outetts docker-build-diffusers docker-build-kokoro docker-build-faster-whisper docker-build-crispasr docker-build-coqui docker-build-chatterbox docker-build-vibevoice docker-build-liquid-audio docker-build-moonshine docker-build-pocket-tts docker-build-qwen-tts docker-build-fish-speech docker-build-faster-qwen3-tts docker-build-qwen-asr docker-build-nemo docker-build-voxcpm docker-build-whisperx docker-build-ace-step docker-build-acestep-cpp docker-build-voxtral docker-build-mlx-distributed docker-build-trl docker-build-llama-cpp-quantization docker-build-tinygrad docker-build-kokoros docker-build-sam3-cpp docker-build-rfdetr-cpp docker-build-qwen3-tts-cpp docker-build-moss-tts-cpp docker-build-magpie-tts-cpp docker-build-vllm-cpp docker-build-omnivoice-cpp docker-build-vibevoice-cpp docker-build-localvqe docker-build-insightface docker-build-speaker-recognition docker-build-sherpa-onnx docker-build-cloud-proxy docker-build-supertonic docker-build-depth-anything-cpp docker-build-moss-transcribe-cpp docker-build-nemo-speech-cpp docker-build-privacy-filter docker-build-trellis2cpp docker-build-valkey-store docker-build-audio-cpp
+docker-build-backends: docker-build-rocmfp4 docker-build-llama-cpp docker-build-ik-llama-cpp docker-build-turboquant docker-build-bonsai docker-build-ds4 docker-build-rerankers docker-build-vllm docker-build-vllm-omni docker-build-longcat-video docker-build-sglang docker-build-transformers docker-build-outetts docker-build-diffusers docker-build-kokoro docker-build-faster-whisper docker-build-crispasr docker-build-coqui docker-build-chatterbox docker-build-vibevoice docker-build-liquid-audio docker-build-moonshine docker-build-pocket-tts docker-build-qwen-tts docker-build-fish-speech docker-build-faster-qwen3-tts docker-build-qwen-asr docker-build-nemo docker-build-voxcpm docker-build-whisperx docker-build-ace-step docker-build-acestep-cpp docker-build-voxtral docker-build-mlx-distributed docker-build-trl docker-build-llama-cpp-quantization docker-build-tinygrad docker-build-kokoros docker-build-sam3-cpp docker-build-rfdetr-cpp docker-build-qwen3-tts-cpp docker-build-moss-tts-cpp docker-build-magpie-tts-cpp docker-build-vllm-cpp docker-build-omnivoice-cpp docker-build-vibevoice-cpp docker-build-localvqe docker-build-insightface docker-build-speaker-recognition docker-build-sherpa-onnx docker-build-cloud-proxy docker-build-supertonic docker-build-depth-anything-cpp docker-build-moss-transcribe-cpp docker-build-nemo-speech-cpp docker-build-privacy-filter docker-build-trellis2cpp docker-build-valkey-store docker-build-audio-cpp
 
 ########################################################
 ### Mock Backend for E2E Tests
diff --git a/backend/Dockerfile.rocmfp4 b/backend/Dockerfile.rocmfp4
new file mode 100644
index 000000000000..7beb88e97685
--- /dev/null
+++ b/backend/Dockerfile.rocmfp4
@@ -0,0 +1,172 @@
+ARG BASE_IMAGE=ubuntu:24.04
+# BUILDER_BASE_IMAGE defaults to BASE_IMAGE so the Dockerfile parses even
+# when no prebuilt base is supplied. The builder-prebuilt stage is only
+# entered when BUILDER_TARGET=builder-prebuilt, so a "wrong" fallback
+# content here is harmless — BuildKit prunes the unreferenced builder.
+ARG BUILDER_BASE_IMAGE=${BASE_IMAGE}
+# BUILDER_TARGET selects which builder stage the final scratch image copies
+# package output from. Declared at global scope (before any FROM) so it's
+# usable in `FROM ${BUILDER_TARGET}` below. Default keeps local
+# `make backends/rocmfp4` on the from-source path.
+ARG BUILDER_TARGET=builder-fromsource
+ARG APT_MIRROR=""
+ARG APT_PORTS_MIRROR=""
+
+
+# ============================================================================
+# Stage: builder-fromsource — self-contained build path.
+# Runs .docker/install-base-deps.sh (apt deps + cmake + protoc + gRPC +
+# conditional CUDA/ROCm/Vulkan), copies /opt/grpc to /usr/local, then
+# compiles the variant. Used when BUILDER_TARGET=builder-fromsource (the
+# default; local `make backends/rocmfp4`).
+#
+# The install script is the same one that backend/Dockerfile.base-grpc-builder
+# runs, so the result is bit-equivalent to the prebuilt-base path
+# (builder-prebuilt below).
+# ============================================================================
+FROM ${BASE_IMAGE} AS builder-fromsource
+ARG BUILD_TYPE
+ARG CUDA_MAJOR_VERSION
+ARG CUDA_MINOR_VERSION
+ARG CMAKE_FROM_SOURCE=false
+# CUDA Toolkit 13.x compatibility: CMake 3.31.9+ fixes toolchain detection/arch table issues
+ARG CMAKE_VERSION=3.31.10
+ARG GRPC_VERSION=v1.65.0
+ARG GRPC_MAKEFLAGS="-j4 -Otarget"
+ARG SKIP_DRIVERS=false
+ARG TARGETARCH
+ARG TARGETVARIANT
+ARG GO_VERSION=1.25.4
+ARG UBUNTU_VERSION=2404
+ARG APT_MIRROR
+ARG APT_PORTS_MIRROR
+ARG AMDGPU_TARGETS=""
+ARG BACKEND=rerankers
+# Local-build escape hatch: point the fetch at a git daemon on the host instead of
+# GitHub. Defaults are empty, so CI keeps using the Makefile's pinned repo+sha.
+ARG LLAMA_REPO=
+ARG ROCMFP4_VERSION=
+ENV LLAMA_REPO=${LLAMA_REPO}
+ENV ROCMFP4_VERSION=${ROCMFP4_VERSION}
+# Local-build escape hatch: point the fetch at a git daemon on the host instead of
+# GitHub. Defaults are empty, so CI keeps using the Makefile's pinned repo+sha.
+ARG LLAMA_REPO=
+ARG ROCMFP4_VERSION=
+ENV LLAMA_REPO=${LLAMA_REPO}
+ENV ROCMFP4_VERSION=${ROCMFP4_VERSION}
+# CUDA target archs, e.g. --build-arg CUDA_DOCKER_ARCH='75;86;89;120'
+ARG CUDA_DOCKER_ARCH
+ARG CMAKE_ARGS
+
+ENV BUILD_TYPE=${BUILD_TYPE} \
+    CUDA_MAJOR_VERSION=${CUDA_MAJOR_VERSION} \
+    CUDA_MINOR_VERSION=${CUDA_MINOR_VERSION} \
+    CMAKE_FROM_SOURCE=${CMAKE_FROM_SOURCE} \
+    CMAKE_VERSION=${CMAKE_VERSION} \
+    GRPC_VERSION=${GRPC_VERSION} \
+    GRPC_MAKEFLAGS=${GRPC_MAKEFLAGS} \
+    SKIP_DRIVERS=${SKIP_DRIVERS} \
+    TARGETARCH=${TARGETARCH} \
+    UBUNTU_VERSION=${UBUNTU_VERSION} \
+    APT_MIRROR=${APT_MIRROR} \
+    APT_PORTS_MIRROR=${APT_PORTS_MIRROR} \
+    AMDGPU_TARGETS=${AMDGPU_TARGETS} \
+    CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH} \
+    CMAKE_ARGS=${CMAKE_ARGS} \
+    DEBIAN_FRONTEND=noninteractive
+
+# CUDA on PATH (no-op when CUDA isn't installed)
+ENV PATH=/usr/local/cuda/bin:${PATH}
+# HipBLAS / ROCm on PATH (no-op when ROCm isn't installed)
+ENV PATH=/opt/rocm/bin:${PATH}
+
+WORKDIR /build
+
+# Install everything via the shared script — the same one that
+# backend/Dockerfile.base-grpc-builder runs, so the prebuilt CI base and
+# this from-source path are bit-equivalent.
+RUN --mount=type=bind,source=.docker/install-base-deps.sh,target=/usr/local/sbin/install-base-deps \
+    --mount=type=bind,source=.docker/apt-mirror.sh,target=/usr/local/sbin/apt-mirror \
+    bash /usr/local/sbin/install-base-deps
+
+# Mirror builder-prebuilt: copy gRPC from /opt/grpc to /usr/local so
+# CMake's find_package finds it at the canonical prefix the Makefile expects.
+RUN cp -a /opt/grpc/. /usr/local/
+
+COPY . /LocalAI
+
+# BuildKit cache mount for ccache. See Dockerfile.llama-cpp (commit 9228e5b4)
+# for rationale. rocmfp4 is a llama.cpp fork that reuses
+# backend/cpp/llama-cpp source via a thin wrapper Makefile, so MOST TUs
+# are content-identical to the upstream llama-cpp build. Sharing a cache
+# id with llama-cpp could give cross-fork hits — but for now keep them
+# separate so a regression in one doesn't poison the other. Revisit
+# sharing after measuring the actual hit rate.
+#
+# The compile body is shared with builder-prebuilt via .docker/rocmfp4-compile.sh.
+RUN --mount=type=bind,source=.docker/rocmfp4-compile.sh,target=/usr/local/sbin/compile.sh \
+    --mount=type=cache,target=/root/.ccache,id=rocmfp4-ccache-${TARGETARCH}-${BUILD_TYPE},sharing=locked \
+    bash /usr/local/sbin/compile.sh
+
+
+# Copy libraries using a script to handle architecture differences
+RUN make -BC /LocalAI/backend/cpp/rocmfp4 package
+
+
+# ============================================================================
+# Stage: builder-prebuilt — uses the pre-built base from
+# quay.io/go-skynet/ci-cache:base-grpc-* (built by .github/workflows/base-images.yml).
+# That image already has gRPC at /opt/grpc + apt deps + CUDA/ROCm/Vulkan
+# pre-installed, so we just copy gRPC to /usr/local and compile. Used when
+# BUILDER_TARGET=builder-prebuilt (CI when the matrix entry sets
+# builder-base-image).
+# ============================================================================
+FROM ${BUILDER_BASE_IMAGE} AS builder-prebuilt
+
+ARG BUILD_TYPE
+ENV BUILD_TYPE=${BUILD_TYPE}
+ARG CUDA_DOCKER_ARCH
+ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
+ARG CMAKE_ARGS
+ENV CMAKE_ARGS=${CMAKE_ARGS}
+# AMDGPU_TARGETS must be forwarded into the env here too — backend/cpp/llama-cpp/Makefile
+# (which the rocmfp4 Makefile reuses via a sibling build dir) errors out when the var
+# is empty on a hipblas build, and the prebuilt path is what CI exercises most of the
+# time. The builder-fromsource stage above already does this; mirror it here.
+ARG AMDGPU_TARGETS
+ENV AMDGPU_TARGETS=${AMDGPU_TARGETS}
+ARG TARGETARCH
+ARG TARGETVARIANT
+
+# The base-grpc-* image installs gRPC to /opt/grpc but doesn't copy it to
+# /usr/local. Mirror what the from-source path does so the compile step
+# can find gRPC at the canonical prefix the Makefile expects.
+RUN cp -a /opt/grpc/. /usr/local/
+
+COPY . /LocalAI
+
+RUN --mount=type=bind,source=.docker/rocmfp4-compile.sh,target=/usr/local/sbin/compile.sh \
+    --mount=type=cache,target=/root/.ccache,id=rocmfp4-ccache-${TARGETARCH}-${BUILD_TYPE},sharing=locked \
+    bash /usr/local/sbin/compile.sh
+
+RUN make -BC /LocalAI/backend/cpp/rocmfp4 package
+
+
+# ============================================================================
+# Final stage — copies package output from one of the two builders.
+# BUILDER_TARGET selects which one. BuildKit prunes the unreferenced builder.
+#
+# BuildKit doesn't support variable expansion in `COPY --from=` directly,
+# so we resolve the ARG by aliasing the chosen builder to a fixed stage
+# name via `FROM ${BUILDER_TARGET} AS builder` and then COPY --from=builder.
+# BUILDER_TARGET itself is declared as a global ARG at the top of this
+# file (required for use in FROM), so we just re-import it into this
+# stage's scope before the FROM directive.
+# ============================================================================
+FROM ${BUILDER_TARGET} AS builder
+
+FROM scratch
+
+
+# Copy all available binaries (the build process only creates the appropriate ones for the target architecture)
+COPY --from=builder /LocalAI/backend/cpp/rocmfp4/package/. ./
diff --git a/backend/cpp/rocmfp4/Makefile b/backend/cpp/rocmfp4/Makefile
new file mode 100644
index 000000000000..5f9db1ef9197
--- /dev/null
+++ b/backend/cpp/rocmfp4/Makefile
@@ -0,0 +1,99 @@
+# Pinned to the HEAD of the `rocmfp4` branch on https://github.com/walcz-de/llama.cpp-ROCmFP4.
+# Auto-bumped nightly by .github/workflows/bump_deps.yaml.
+ROCMFP4_VERSION?=0b6afd26bfa64f51c14794b4dfbe3da8946076d3
+LLAMA_REPO?=https://github.com/walcz-de/llama.cpp-ROCmFP4
+
+CMAKE_ARGS?=
+BUILD_TYPE?=
+NATIVE?=false
+ONEAPI_VARS?=/opt/intel/oneapi/setvars.sh
+TARGET?=--target grpc-server
+JOBS?=$(shell nproc 2>/dev/null || sysctl -n hw.ncpu 2>/dev/null || echo 1)
+ARCH?=$(shell uname -m)
+
+CURRENT_MAKEFILE_DIR := $(dir $(abspath $(lastword $(MAKEFILE_LIST))))
+LLAMA_CPP_DIR := $(CURRENT_MAKEFILE_DIR)/../llama-cpp
+
+GREEN := \033[0;32m
+RESET := \033[0m
+
+# rocmfp4 is a llama.cpp fork carrying the ROCmFP4 / ROCmFPx weight-quantization
+# formats (ggml types 100-107) for AMD RDNA3.5 APUs. Stock llama.cpp rejects those
+# tensor types, which is the whole reason this backend exists as a separate image
+# rather than a flag on llama-cpp.
+#
+# Like bonsai, the additions are model *weight* types decoded inside libllama and
+# therefore transparent to the shared gRPC server: no grpc-server.cpp allow-list
+# patch is needed (unlike turboquant, whose KV-cache types do need one). So we reuse
+# backend/cpp/llama-cpp's grpc-server.cpp / CMakeLists.txt / Makefile verbatim and
+# only swap which repo and commit the fetch step pulls.
+PATCHES_DIR := $(CURRENT_MAKEFILE_DIR)/patches
+
+define rocmfp4-build
+	rm -rf $(CURRENT_MAKEFILE_DIR)/../rocmfp4-$(1)-build
+	cp -rf $(LLAMA_CPP_DIR) $(CURRENT_MAKEFILE_DIR)/../rocmfp4-$(1)-build
+	# Drop patches vendored for upstream llama.cpp: this fork carries them already
+	# (it is based on the same pin), so re-applying them would reject.
+	rm -rf $(CURRENT_MAKEFILE_DIR)/../rocmfp4-$(1)-build/patches
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../rocmfp4-$(1)-build purge
+	bash $(LLAMA_CPP_DIR)/disable-score-task.sh $(CURRENT_MAKEFILE_DIR)/../rocmfp4-$(1)-build/grpc-server.cpp
+	bash $(LLAMA_CPP_DIR)/disable-tts-task.sh $(CURRENT_MAKEFILE_DIR)/../rocmfp4-$(1)-build/grpc-server.cpp
+	$(info $(GREEN)I rocmfp4 build info:$(1)$(RESET))
+	LLAMA_REPO=$(LLAMA_REPO) LLAMA_VERSION=$(ROCMFP4_VERSION) \
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../rocmfp4-$(1)-build llama.cpp
+	bash $(CURRENT_MAKEFILE_DIR)/apply-patches.sh $(CURRENT_MAKEFILE_DIR)/../rocmfp4-$(1)-build/llama.cpp $(PATCHES_DIR)
+	CMAKE_ARGS="$(CMAKE_ARGS) $(2)" TARGET="$(3)" \
+	LLAMA_REPO=$(LLAMA_REPO) LLAMA_VERSION=$(ROCMFP4_VERSION) \
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../rocmfp4-$(1)-build grpc-server
+	cp -rfv $(CURRENT_MAKEFILE_DIR)/../rocmfp4-$(1)-build/grpc-server rocmfp4-$(1)
+endef
+
+rocmfp4-avx2:
+	$(call rocmfp4-build,avx2,-DGGML_AVX=on -DGGML_AVX2=on -DGGML_AVX512=off -DGGML_FMA=on -DGGML_F16C=on,--target grpc-server)
+
+rocmfp4-avx512:
+	$(call rocmfp4-build,avx512,-DGGML_AVX=on -DGGML_AVX2=off -DGGML_AVX512=on -DGGML_FMA=on -DGGML_F16C=on,--target grpc-server)
+
+rocmfp4-avx:
+	$(call rocmfp4-build,avx,-DGGML_AVX=on -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off,--target grpc-server)
+
+rocmfp4-fallback:
+	$(call rocmfp4-build,fallback,-DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off,--target grpc-server)
+
+# Single-build CPU backend via ggml CPU_ALL_VARIANTS (mirrors llama-cpp-cpu-all).
+rocmfp4-cpu-all:
+	rm -rf $(CURRENT_MAKEFILE_DIR)/../rocmfp4-cpu-all-build
+	cp -rf $(LLAMA_CPP_DIR) $(CURRENT_MAKEFILE_DIR)/../rocmfp4-cpu-all-build
+	rm -rf $(CURRENT_MAKEFILE_DIR)/../rocmfp4-cpu-all-build/patches
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../rocmfp4-cpu-all-build purge
+	bash $(LLAMA_CPP_DIR)/disable-score-task.sh $(CURRENT_MAKEFILE_DIR)/../rocmfp4-cpu-all-build/grpc-server.cpp
+	bash $(LLAMA_CPP_DIR)/disable-tts-task.sh $(CURRENT_MAKEFILE_DIR)/../rocmfp4-cpu-all-build/grpc-server.cpp
+	$(info $(GREEN)I rocmfp4 build info:cpu-all-variants$(RESET))
+	LLAMA_REPO=$(LLAMA_REPO) LLAMA_VERSION=$(ROCMFP4_VERSION) \
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../rocmfp4-cpu-all-build llama.cpp
+	bash $(CURRENT_MAKEFILE_DIR)/apply-patches.sh $(CURRENT_MAKEFILE_DIR)/../rocmfp4-cpu-all-build/llama.cpp $(PATCHES_DIR)
+	SHARED_LIBS=ON EXTRA_CMAKE_ARGS="-DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON" TARGET="--target grpc-server --target ggml" \
+	LLAMA_REPO=$(LLAMA_REPO) LLAMA_VERSION=$(ROCMFP4_VERSION) \
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../rocmfp4-cpu-all-build grpc-server
+	cp -rfv $(CURRENT_MAKEFILE_DIR)/../rocmfp4-cpu-all-build/grpc-server rocmfp4-cpu-all
+	rm -rf ggml-shared-libs && mkdir -p ggml-shared-libs
+	find $(CURRENT_MAKEFILE_DIR)/../rocmfp4-cpu-all-build/llama.cpp/build \( -name '*.so*' -o -name '*.dylib' \) -exec cp -av {} ggml-shared-libs/ \;
+	@echo "Collected ggml shared backends:" && ls -la ggml-shared-libs/
+
+rocmfp4-grpc:
+	$(call rocmfp4-build,grpc,-DGGML_RPC=ON -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off,--target grpc-server --target ggml-rpc-server)
+
+# Upstream renamed the RPC binary rpc-server -> ggml-rpc-server. The bonsai wrapper
+# this was modelled on still uses the old name because its fork predates the rename;
+# ours is based on a current tree, so it must use the new one.
+rocmfp4-rpc-server: rocmfp4-grpc
+	cp -rf $(CURRENT_MAKEFILE_DIR)/../rocmfp4-grpc-build/llama.cpp/build/bin/ggml-rpc-server rocmfp4-rpc-server
+
+package:
+	bash package.sh
+
+purge:
+	rm -rf $(CURRENT_MAKEFILE_DIR)/../rocmfp4-*-build
+	rm -rf rocmfp4-* package
+
+clean: purge
diff --git a/backend/cpp/rocmfp4/apply-patches.sh b/backend/cpp/rocmfp4/apply-patches.sh
new file mode 100755
index 000000000000..bdee6ba0627a
--- /dev/null
+++ b/backend/cpp/rocmfp4/apply-patches.sh
@@ -0,0 +1,50 @@
+#!/bin/bash
+# Apply the rocmfp4 patch series to a cloned walcz-de/llama.cpp-ROCmFP4 checkout.
+#
+# The rocmfp4 fork is rebased onto the same LLAMA_VERSION LocalAI pins for the shared
+# backend/cpp/llama-cpp/grpc-server.cpp, so in the common case patches/ is empty and this
+# script is a no-op. If the pins ever drift — upstream changes an API the shared gRPC
+# server depends on before the fork has rebased — the gap is carried as patch files under
+# backend/cpp/rocmfp4/patches/ and applied here so the reused grpc-server source compiles
+# against the fork unmodified.
+#
+# Drop the corresponding patch from patches/ whenever the fork catches up with upstream —
+# the build will fail fast if a patch stops applying, which is the signal to retire it.
+
+set -euo pipefail
+
+if [[ $# -ne 2 ]]; then
+    echo "usage: $0 <llama.cpp-src-dir> <patches-dir>" >&2
+    exit 2
+fi
+
+SRC_DIR=$1
+PATCHES_DIR=$2
+
+if [[ ! -d "$SRC_DIR" ]]; then
+    echo "source dir does not exist: $SRC_DIR" >&2
+    exit 2
+fi
+
+if [[ ! -d "$PATCHES_DIR" ]]; then
+    echo "no patches dir at $PATCHES_DIR, nothing to apply"
+    exit 0
+fi
+
+shopt -s nullglob
+patches=("$PATCHES_DIR"/*.patch)
+shopt -u nullglob
+
+if [[ ${#patches[@]} -eq 0 ]]; then
+    echo "no .patch files in $PATCHES_DIR, nothing to apply"
+    exit 0
+fi
+
+cd "$SRC_DIR"
+
+for patch in "${patches[@]}"; do
+    echo "==> applying $patch"
+    git apply --verbose "$patch"
+done
+
+echo "all rocmfp4 patches applied successfully"
diff --git a/backend/cpp/rocmfp4/package.sh b/backend/cpp/rocmfp4/package.sh
new file mode 100755
index 000000000000..691c1abd801a
--- /dev/null
+++ b/backend/cpp/rocmfp4/package.sh
@@ -0,0 +1,39 @@
+#!/bin/bash
+
+# Script to copy the appropriate libraries based on architecture
+# This script is used in the final stage of the Dockerfile
+
+set -e
+
+CURDIR=$(dirname "$(realpath $0)")
+REPO_ROOT="${CURDIR}/../../.."
+
+# Create lib directory
+mkdir -p $CURDIR/package/lib
+
+cp -avrf $CURDIR/rocmfp4-* $CURDIR/package/
+cp -rfv $CURDIR/run.sh $CURDIR/package/
+
+# Bundle the ggml shared backends from the CPU_ALL_VARIANTS build into package/lib. ggml
+# discovers the per-microarch libggml-cpu-*.so by scanning the executable directory, which
+# (via the bundled lib/ld.so that run.sh launches through) resolves to lib/. See the
+# matching comment in backend/cpp/llama-cpp/package.sh. No-op on the fallback/ROCm builds.
+if [ -d "$CURDIR/ggml-shared-libs" ]; then
+    echo "Bundling ggml shared backends (CPU_ALL_VARIANTS)..."
+    cp -avf $CURDIR/ggml-shared-libs/*.so* $CURDIR/package/lib/
+fi
+
+# Detect architecture and copy appropriate libraries
+source "$CURDIR/../../../scripts/build/package-system-libs.sh" "$CURDIR/package/lib" ""
+
+# Package GPU libraries based on BUILD_TYPE
+GPU_LIB_SCRIPT="${REPO_ROOT}/scripts/build/package-gpu-libs.sh"
+if [ -f "$GPU_LIB_SCRIPT" ]; then
+    echo "Packaging GPU libraries for BUILD_TYPE=${BUILD_TYPE:-cpu}..."
+    source "$GPU_LIB_SCRIPT" "$CURDIR/package/lib"
+    package_gpu_libs
+fi
+
+echo "Packaging completed successfully"
+ls -liah $CURDIR/package/
+ls -liah $CURDIR/package/lib/
diff --git a/backend/cpp/rocmfp4/patches/README.md b/backend/cpp/rocmfp4/patches/README.md
new file mode 100644
index 000000000000..e56c93c7000e
--- /dev/null
+++ b/backend/cpp/rocmfp4/patches/README.md
@@ -0,0 +1,15 @@
+# rocmfp4 fork skew patches
+
+The `rocmfp4` backend reuses `backend/cpp/llama-cpp/grpc-server.cpp` (written against
+LocalAI's pinned *upstream* llama.cpp) but compiles it against the
+[walcz-de/llama.cpp-ROCmFP4](https://github.com/walcz-de/llama.cpp-ROCmFP4) fork, which
+carries the ROCmFP4 / ROCmFPx tensor formats (ggml types 100-107) for AMD RDNA3.5 APUs.
+
+The fork tracks the same upstream pin as `backend/cpp/llama-cpp` (it is re-based onto the
+`LLAMA_VERSION` LocalAI pins), so in the common case **no patches are needed here** and this
+directory stays empty. If the pins ever drift — upstream changes an API the shared gRPC
+server depends on before the fork has rebased — the gap is back-ported here as a `*.patch`
+file and applied to the cloned fork checkout by `../apply-patches.sh`.
+
+Patches apply with `git apply` from the fork checkout root. Name them
+`NNNN-short-description.patch` so the apply order stays deterministic.
diff --git a/backend/cpp/rocmfp4/run.sh b/backend/cpp/rocmfp4/run.sh
new file mode 100755
index 000000000000..0ccf3cae7137
--- /dev/null
+++ b/backend/cpp/rocmfp4/run.sh
@@ -0,0 +1,77 @@
+#!/bin/bash
+set -ex
+
+# Get the absolute current dir where the script is located
+CURDIR=$(dirname "$(realpath "$0")")
+
+cd /
+
+echo "CPU info:"
+grep -e "model\sname" /proc/cpuinfo | head -1
+grep -e "flags" /proc/cpuinfo | head -1
+
+BINARY=rocmfp4-fallback
+
+# x86/arm64 ship a single rocmfp4-cpu-all built with ggml CPU_ALL_VARIANTS: ggml's
+# backend registry dlopens the best libggml-cpu-*.so for this host, so no shell-side
+# probing. ROCm ships only rocmfp4-fallback, so fall back to it when cpu-all is absent.
+if [ -e "$CURDIR"/rocmfp4-cpu-all ]; then
+	BINARY=rocmfp4-cpu-all
+fi
+
+if [ -n "$LLAMACPP_GRPC_SERVERS" ]; then
+	if [ -e "$CURDIR"/rocmfp4-grpc ]; then
+		BINARY=rocmfp4-grpc
+	fi
+fi
+
+# Extend ld library path with the dir where this script is located/lib
+if [ "$(uname)" == "Darwin" ]; then
+	export DYLD_LIBRARY_PATH="$CURDIR"/lib:$DYLD_LIBRARY_PATH
+else
+	export LD_LIBRARY_PATH="$CURDIR"/lib:$LD_LIBRARY_PATH
+	# Tell rocBLAS where to find TensileLibrary data (GPU kernel tuning files)
+	if [ -d "$CURDIR/lib/rocblas/library" ]; then
+		export ROCBLAS_TENSILE_LIBPATH="$CURDIR"/lib/rocblas/library
+	fi
+	# Same for hipBLASLt (rocblaslt): the bundled libhipblaslt.so resolves its
+	# TensileLibrary_lazy_gfx*.dat kernel data relative to itself, so point it at
+	# the bundled data or it falls back to slow generic kernels (issue #10660).
+	if [ -d "$CURDIR/lib/hipblaslt/library" ]; then
+		export HIPBLASLT_TENSILE_LIBPATH="$CURDIR"/lib/hipblaslt/library
+	fi
+	# Backends built for Intel GPUs carry a copy of the Intel graphics driver,
+	# and libze_loader is only there in those builds. Level Zero looks for a
+	# driver on its own, so point it at the copy that came with this backend: it
+	# was built against the same C library, while the machine's own driver may
+	# not have been, and loading that one can crash on start.
+	#
+	# Anything the user set is left alone, so a machine with a graphics card
+	# newer than the driver carried here can still be told to use its own.
+	# Nothing is said about OpenCL: no OpenCL driver is carried, so anything we
+	# set there would leave OpenCL worse off than the machine's own setup.
+	if [ -e "$CURDIR/lib/libze_loader.so.1" ]; then
+		if [ -e "$CURDIR/lib/libze_intel_gpu.so.1" ] && [ -z "${ZE_ENABLE_ALT_DRIVERS:-}" ]; then
+			export ZE_ENABLE_ALT_DRIVERS="$CURDIR"/lib/libze_intel_gpu.so.1
+		fi
+		# Ask the driver how much graphics memory is free. Without this, the
+		# backend reads zero on an integrated graphics chip, because such a chip
+		# shares the system memory instead of having its own.
+		if [ -z "${ZES_ENABLE_SYSMAN:-}" ]; then
+			export ZES_ENABLE_SYSMAN=1
+		fi
+	fi
+fi
+
+# If there is a lib/ld.so, use it
+if [ -f "$CURDIR"/lib/ld.so ]; then
+	echo "Using lib/ld.so"
+	echo "Using binary: $BINARY"
+	exec "$CURDIR"/lib/ld.so "$CURDIR"/$BINARY "$@"
+fi
+
+echo "Using binary: $BINARY"
+exec "$CURDIR"/$BINARY "$@"
+
+# We should never reach this point, however just in case we do, run fallback
+exec "$CURDIR"/rocmfp4-fallback "$@"
diff --git a/backend/index.yaml b/backend/index.yaml
index 3420d85496b7..e17431b7c901 100644
--- a/backend/index.yaml
+++ b/backend/index.yaml
@@ -74,6 +74,43 @@
     nvidia-cuda-12: "cuda12-turboquant"
     nvidia-l4t-cuda-12: "nvidia-l4t-arm64-turboquant"
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-turboquant"
+- &rocmfp4
+  name: "rocmfp4"
+  alias: "rocmfp4"
+  license: mit
+  description: |
+    Fork of llama.cpp carrying the ROCmFP4 / ROCmFPx 4-bit weight-quantization
+    formats (ggml types 100-107) for AMD RDNA3.5 APUs - gfx1150/1151/1152/1153,
+    which is Strix Point, Strix Halo, Gorgon Point and Gorgon Halo. Stock
+    llama.cpp rejects these tensor types, so this backend exists to serve the
+    published ROCmFP4 weights rather than to be a faster llama.cpp.
+
+    What the format buys on this hardware is MEMORY, not arithmetic: roughly 22%
+    smaller than Q4_K_M at perplexity parity, which can decide whether a model
+    fits in unified RAM. There is no FP4 matrix instruction on RDNA3.5 or RDNA4 -
+    only CDNA4 has one - so the matmul runs as int8 dot products and the gain is
+    bandwidth. Measured on a Radeon 8060S (gfx1151): 335 tok/s prefill and
+    14.2 tok/s decode on a 27B model.
+
+    llama-cpp stays the recommendation for everything else.
+  urls:
+    - https://github.com/walcz-de/llama.cpp-ROCmFP4
+    - https://github.com/charlie12345/ROCmFPX
+  tags:
+    - text-to-text
+    - LLM
+    - GPU
+    - HIP
+    - AMD
+    - rocmfp4
+    - fp4
+    - strix-halo
+  capabilities:
+    amd: "rocm-rocmfp4"
+- !!merge <<: *rocmfp4
+  name: "rocmfp4-development"
+  capabilities:
+    amd: "rocm-rocmfp4-development"
 - &bonsai
   name: "bonsai"
   alias: "bonsai"
@@ -2859,6 +2896,16 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-13-bonsai"
   mirrors:
     - localai/localai-backends:master-gpu-nvidia-cuda-13-bonsai
+- !!merge <<: *rocmfp4
+  name: "rocm-rocmfp4"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-rocmfp4"
+  mirrors:
+    - localai/localai-backends:latest-gpu-rocm-hipblas-rocmfp4
+- !!merge <<: *rocmfp4
+  name: "rocm-rocmfp4-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-rocmfp4"
+  mirrors:
+    - localai/localai-backends:master-gpu-rocm-hipblas-rocmfp4
 - !!merge <<: *bonsai
   name: "rocm-bonsai"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-bonsai"
diff --git a/core/gallery/importers/llama-cpp.go b/core/gallery/importers/llama-cpp.go
index a1cbb6d1bc3c..bc70de74e790 100644
--- a/core/gallery/importers/llama-cpp.go
+++ b/core/gallery/importers/llama-cpp.go
@@ -37,6 +37,7 @@ func (i *LlamaCPPImporter) AdditionalBackends() []KnownBackendEntry {
 	return []KnownBackendEntry{
 		{Name: "ik-llama-cpp", Modality: "text", Description: "GGUF drop-in replacement for llama-cpp with ik-quants"},
 		{Name: "turboquant", Modality: "text", Description: "GGUF drop-in replacement for llama-cpp with TurboQuant optimizations"},
+		{Name: "rocmfp4", Modality: "text", Description: "GGUF drop-in replacement for llama-cpp reading ROCmFP4 4-bit weights on AMD RDNA3.5"},
 		{Name: "vllm-cpp", Modality: "text", Description: "vLLM-style continuous-batching engine (vllm.cpp) consuming GGUF, by the LocalAI team"},
 	}
 }
@@ -136,7 +137,9 @@ func (i *LlamaCPPImporter) Import(details Details) (gallery.ModelConfig, error)
 	backend := "llama-cpp"
 	if b, ok := preferencesMap["backend"].(string); ok {
 		switch b {
-		case "ik-llama-cpp", "turboquant", "vllm-cpp":
+		// Preference-only: ROCmFP4 GGUFs carry ordinary .gguf names, so there is no
+		// safe auto-detect signal. Without an explicit preference this stays llama-cpp.
+		case "ik-llama-cpp", "turboquant", "vllm-cpp", "rocmfp4":
 			backend = b
 		}
 	}
diff --git a/core/gallery/importers/llama-cpp_test.go b/core/gallery/importers/llama-cpp_test.go
index 139b29884bb7..453035e6d05f 100644
--- a/core/gallery/importers/llama-cpp_test.go
+++ b/core/gallery/importers/llama-cpp_test.go
@@ -181,6 +181,23 @@ var _ = Describe("LlamaCPPImporter", func() {
 			Expect(modelConfig.Files[0].Filename).To(Equal("my-model.gguf"))
 		})
 
+		It("swaps the emitted backend to rocmfp4 when preferred", func() {
+			preferences := json.RawMessage(`{"backend": "rocmfp4"}`)
+			details := Details{
+				URI:         "https://example.com/my-model.gguf",
+				Preferences: preferences,
+			}
+
+			modelConfig, err := importer.Import(details)
+
+			Expect(err).ToNot(HaveOccurred())
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("backend: rocmfp4"), fmt.Sprintf("Model config: %+v", modelConfig))
+			Expect(modelConfig.ConfigFile).NotTo(ContainSubstring("backend: llama-cpp\n"), fmt.Sprintf("Model config: %+v", modelConfig))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("model: my-model.gguf"), fmt.Sprintf("Model config: %+v", modelConfig))
+			Expect(len(modelConfig.Files)).To(Equal(1))
+			Expect(modelConfig.Files[0].Filename).To(Equal("my-model.gguf"))
+		})
+
 		It("swaps the emitted backend to vllm-cpp when preferred, keeping engine-side templating", func() {
 			preferences := json.RawMessage(`{"backend": "vllm-cpp"}`)
 			details := Details{
diff --git a/docs/content/features/backends.md b/docs/content/features/backends.md
index 7d2ec66a4f4a..a18487780a73 100644
--- a/docs/content/features/backends.md
+++ b/docs/content/features/backends.md
@@ -174,6 +174,7 @@ For getting started, see the available backends in LocalAI here: https://github.
 LocalAI supports various types of backends:
 
 - **LLM Backends**: For running language models (e.g., llama.cpp, vLLM, vllm.cpp, SGLang, transformers, MLX)
+- **AMD FP4 Backend**: `rocmfp4` — a llama.cpp fork reading the ROCmFP4 / ROCmFPx 4-bit weight formats (ggml types 100-107) on AMD RDNA3.5 APUs (Strix Point / Strix Halo / Gorgon). Stock llama.cpp rejects these tensor types. The format trades memory, not arithmetic: ~22% smaller than Q4_K_M at perplexity parity, with the matmul running as int8 dot products since RDNA3.5 has no FP4 matrix instruction. `llama-cpp` remains the recommendation for everything else.
 - **Speech-to-Text Backends**: For transcription, forced alignment and speaker diarization (e.g., whisper.cpp, parakeet.cpp, moss-transcribe.cpp, [NeMo-Speech.cpp]({{%relref "features/nemo-speech-cpp" %}}), faster-whisper, NeMo, [audio.cpp]({{%relref "features/audio-cpp" %}}))
 - **Text-to-Speech Backends**: For speech synthesis (e.g., piper, Kokoro, VibeVoice, Qwen3-TTS, [NeMo-Speech.cpp]({{%relref "features/nemo-speech-cpp" %}}), [audio.cpp]({{%relref "features/audio-cpp" %}}))
 - **Sound Generation Backends**: For music and audio generation (e.g., ACE-Step, [audio.cpp]({{%relref "features/audio-cpp" %}}))
diff --git a/scripts/lib/backend-filter.mjs b/scripts/lib/backend-filter.mjs
index 9312cbf6d0c0..6c6e976a7092 100644
--- a/scripts/lib/backend-filter.mjs
+++ b/scripts/lib/backend-filter.mjs
@@ -78,6 +78,12 @@ export function inferBackendPath(item) {
     // via a thin wrapper Makefile. Changes to either dir should retrigger it.
     return `backend/cpp/turboquant/`;
   }
+  if (item.dockerfile.endsWith("rocmfp4")) {
+    // rocmfp4 is a llama.cpp fork carrying the ROCmFP4 weight formats; like
+    // bonsai it reuses backend/cpp/llama-cpp sources via a wrapper Makefile.
+    // Must stay above the generic llama-cpp suffix test.
+    return `backend/cpp/rocmfp4/`;
+  }
   if (item.dockerfile.endsWith("bonsai")) {
     // bonsai is a llama.cpp fork that reuses backend/cpp/llama-cpp sources
     // via a thin wrapper Makefile. Changes to either dir should retrigger it.
@@ -152,7 +158,7 @@ export function backendChanged(backend, pathPrefix, changedFiles) {
 
   // Fork backends reuse backend/cpp/llama-cpp sources via thin wrappers;
   // changes to either directory must retrigger their pipelines.
-  return (backend === "turboquant" || backend === "bonsai") &&
+  return (backend === "turboquant" || backend === "bonsai" || backend === "rocmfp4") &&
     changedFiles.some(file => file.startsWith("backend/cpp/llama-cpp/"));
 }