diff --git a/.SRCINFO b/.SRCINFO
index 2ad6932bbe760..4e42304cb7055 100644
--- a/.SRCINFO
+++ b/.SRCINFO
@@ -1,6 +1,6 @@
 pkgbase = llama.cpp-hip
 	pkgdesc = Port of Facebook's LLaMA model in C/C++ (with AMD ROCm optimizations)
-	pkgver = b6834
+	pkgver = b6838
 	pkgrel = 1
 	url = https://github.com/ggml-org/llama.cpp
 	arch = x86_64
@@ -31,7 +31,7 @@ pkgbase = llama.cpp-hip
 	conflicts = stable-diffusion.cpp
 	options = lto
 	options = !debug
-	source = llama.cpp-hip-b6834.tar.gz::https://github.com/ggml-org/llama.cpp/archive/refs/tags/b6834.tar.gz
-	sha256sums = 0aa33fbb9c1d44175c5e7b3a34eea3af436fb0afe4a7146cfacbb6af301db2c9
+	source = llama.cpp-hip-b6838.tar.gz::https://github.com/ggml-org/llama.cpp/archive/refs/tags/b6838.tar.gz
+	sha256sums = c5e499f70f90231d08646f6c3161445ca9101b3d05f1c13486909f1dada07912
 
 pkgname = llama.cpp-hip
diff --git a/PKGBUILD b/PKGBUILD
index a373c8ab99d4b..ff05083ef682c 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -3,7 +3,7 @@
 
 pkgname=llama.cpp-hip
 _pkgname="${pkgname%-hip}"
-pkgver=b6834
+pkgver=b6838
 pkgrel=1
 pkgdesc="Port of Facebook's LLaMA model in C/C++ (with AMD ROCm optimizations)"
 arch=(x86_64 armv7h aarch64)
@@ -36,7 +36,7 @@ provides=(${_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=('0aa33fbb9c1d44175c5e7b3a34eea3af436fb0afe4a7146cfacbb6af301db2c9')
+sha256sums=('c5e499f70f90231d08646f6c3161445ca9101b3d05f1c13486909f1dada07912')
 
 prepare() {
   ln -sf "${_pkgname}-${pkgver}" llama.cpp