diff --git a/.SRCINFO b/.SRCINFO
index 292076a061000..f950906e6fe47 100644
--- a/.SRCINFO
+++ b/.SRCINFO
@@ -1,6 +1,6 @@
 pkgbase = llama.cpp-vulkan
 	pkgdesc = Port of Facebook's LLaMA model in C/C++ (with Vulkan GPU optimizations)
-	pkgver = b7120
+	pkgver = b7128
 	pkgrel = 1
 	url = https://github.com/ggerganov/llama.cpp
 	arch = x86_64
@@ -28,7 +28,7 @@ pkgbase = llama.cpp-vulkan
 	conflicts = stable-diffusion.cpp
 	options = lto
 	options = !debug
-	source = llama.cpp-vulkan-b7120.tar.gz::https://github.com/ggml-org/llama.cpp/archive/refs/tags/b7120.tar.gz
-	sha256sums = bea5e98cc756c972a10d6b40cef38fa87190c6acece38960cb1e8ddf4e35c8ae
+	source = llama.cpp-vulkan-b7128.tar.gz::https://github.com/ggml-org/llama.cpp/archive/refs/tags/b7128.tar.gz
+	sha256sums = 1129e48ef64b130f15738257c1c04e9ffa6d0fef39efd30866eaf51e00522f8e
 
 pkgname = llama.cpp-vulkan
diff --git a/PKGBUILD b/PKGBUILD
index 5fe5368d5f105..bd3b6fc36912a 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -3,7 +3,7 @@
 
 pkgname=llama.cpp-vulkan
 _pkgname=${pkgname%%-vulkan}
-pkgver=b7120
+pkgver=b7128
 pkgrel=1
 pkgdesc="Port of Facebook's LLaMA model in C/C++ (with Vulkan GPU optimizations)"
 arch=(x86_64 armv7h aarch64)
@@ -33,7 +33,7 @@ provides=(${_pkgname})
 conflicts=(${_pkgname} libggml ggml stable-diffusion.cpp)
 options=(lto !debug)
 source=("${pkgname}-${pkgver}.tar.gz::https://github.com/ggml-org/llama.cpp/archive/refs/tags/${pkgver}.tar.gz")
-sha256sums=('bea5e98cc756c972a10d6b40cef38fa87190c6acece38960cb1e8ddf4e35c8ae')
+sha256sums=('1129e48ef64b130f15738257c1c04e9ffa6d0fef39efd30866eaf51e00522f8e')
 
 prepare() {
   ln -sf "${_pkgname}-${pkgver}" llama.cpp