Package Details: llama.cpp-cuda b9279-1

Git Clone URL: https://aur.archlinux.org/llama.cpp-cuda.git (read-only, click to copy)
Package Base: llama.cpp-cuda
Description: Port of Facebook's LLaMA model in C/C++ (with NVIDIA CUDA optimizations)
Upstream URL: https://github.com/ggml-org/llama.cpp
Licenses: MIT
Conflicts: ggml, libggml, llama.cpp
Provides: llama.cpp
Submitter: txtsd
Maintainer: fabse
Last Packager: fabse
Votes: 17
Popularity: 1.88
First Submitted: 2024-10-26 20:17 (UTC)
Last Updated: 2026-05-22 06:00 (UTC)

Dependencies (18)

Required by (5)

Sources (3)

Pinned Comments

txtsd commented on 2024-10-26 20:17 (UTC) (edited on 2024-12-06 14:15 (UTC) by txtsd)

Alternate versions

llama.cpp
llama.cpp-vulkan
llama.cpp-sycl-fp16
llama.cpp-sycl-fp32
llama.cpp-cuda
llama.cpp-cuda-f16
llama.cpp-hip

Latest Comments

1 2 3 4 5 6 .. 9 Next › Last »

doaxan commented on 2026-05-20 12:30 (UTC)

Web UI is broken for b9245-1: https://github.com/ggml-org/llama.cpp/issues/23379 https://github.com/ggml-org/llama.cpp/issues/23105

fabse commented on 2026-05-07 18:51 (UTC)

oh right the config should be kept between installs, it will be included in the next release

alegru commented on 2026-05-07 14:47 (UTC)

Can you add backup=("etc/conf.d/llama.cpp") to PKGBUILD?

fabse commented on 2026-05-06 20:09 (UTC)

oh nice, thanks for letting me know, I can confirm the workaround is not needed anymore and is now removed (will be pushed here on the next release)

svenstaro commented on 2026-05-06 17:54 (UTC)

FYI gcc15 is now in [extra] and after re-sourcing your profile.d, $NVCC_CCBIN should correctly point to /usr/bin/g++-15 and all the CUDA stuff should just work.

fabse commented on 2026-05-04 19:50 (UTC) (edited on 2026-05-04 20:38 (UTC) by fabse)

finally, took ages to build the CI-image, but now this package will be automatically updated!

I've added gcc15 as a make dependency, but I'll recommend installing it beforehand to avoid recompiling each time this package gets updated

NOTE: I actually don't use this package myself, so if anyone notices that a new version of cuda with support for gcc16 is released, lemme know and I'll remove the make dependency

fabse commented on 2026-05-04 10:34 (UTC)

@Jark5455 oh geez I thought it wouldn't be more than a week for nvidia to add support, guess I was expecting too much lol

but first I need to create a new CI-image that has gcc15 compiled, then I can push a release here

Jark5455 commented on 2026-05-03 23:36 (UTC)

@fabse we should just add gcc15 as a dependency. gcc16 was just released so it will take nvidia 2-3 months to add support. A lot of other cuda dependent packages (e.g gromacs) just add the older compiler as a dependency.

fabse commented on 2026-05-03 10:37 (UTC) (edited on 2026-05-03 10:39 (UTC) by fabse)

Welp bad timing taking over this package, Arch updated to gcc16 but the cuda-package shipped doesn't support it yet...

EDIT: A temporary solution would be making gcc15 a dependency, but that takes ages to compile

svenstaro commented on 2026-04-27 05:47 (UTC)

@pussyhut: You requested package deletion. There is no reason to delete this package even if the maintainer is missing. You should have requested for the package to be orphaned instead.

At any rate, the package will be eligible for orphaning in 2 days.

1 2 3 4 5 6 .. 9 Next › Last »