{"id":4438,"date":"2026-09-25T22:14:09","date_gmt":"2026-09-25T13:14:09","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/09\/25\/ramalama-v0-25-0-released\/"},"modified":"2026-09-25T22:14:09","modified_gmt":"2026-09-25T13:14:09","slug":"ramalama-v0-25-0-released","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/25\/ramalama-v0-25-0-released\/","title":{"rendered":"RamaLama v0.25.0 Released: Security and Engine Updates"},"content":{"rendered":"

<\/p>\n

At a Glance<\/h2>\n
\n\n\n\n\n\n\n\n\n\n
Item<\/th>\nValue<\/th>\n<\/tr>\n<\/thead>\n
Repository<\/td>\ncontainers\/ramalama<\/a><\/td>\n<\/tr>\n
Version<\/td>\nv0.25.0<\/a><\/td>\n<\/tr>\n
Published<\/td>\n2026-09-25<\/td>\n<\/tr>\n
License<\/td>\nMIT<\/td>\n<\/tr>\n
Source type<\/td>\nPrimary source (the publisher itself)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Values determined by this site’s code at collection time. Dates are JST.<\/em><\/p>\n

<\/p>\n

Overview<\/h2>\n

RamaLama is an open-source Python developer tool that simplifies local serving of AI models from any source and facilitates production inference usage through container mechanisms. This v0.25.0 update brings enhanced network security, updated inference engines, and improved environment compatibility.<\/p>\n

Notably, the change to the default bind setting for the serve<\/code> command is an important update for users exposing models over the network. Additionally, caution is advised for those using existing workflows, as specific API transports have been deprecated.<\/p>\n

Breaking Changes and Deprecations<\/h2>\n

This release includes changes to default values and the deprecation of certain features.<\/p>\n

\n\n\n\n\n\n\n
Item<\/th>\nOld Setting \/ Behavior<\/th>\nNew Setting \/ Behavior<\/th>\n<\/tr>\n<\/thead>\n
serve<\/code> command bind<\/td>\n(Unspecified \/ broad bind)<\/td>\nBinds to loopback by default<\/td>\n<\/tr>\n
hosted API transport<\/td>\nAvailable<\/td>\nDeprecated<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

The change to the serve<\/code> command affects users attempting to access the RamaLama server from other machines on the network. Because it binds to loopback (127.0.0.1) by default, external connections will no longer be accepted unless the setting is explicitly changed.<\/p>\n

Furthermore, the hosted API transport has been deprecated. If you use systems or scripts that rely on this feature, you should consider migrating to other transport methods in preparation for its eventual removal.<\/p>\n

Key Changes<\/h2>\n

Vulkan Support in NVIDIA Environments<\/h3>\n

Changes have been made to enable Vulkan for users utilizing NVIDIA GPUs. This provides new options utilizing Vulkan when running inference in NVIDIA environments.<\/p>\n

llama.cpp Version Upgrade<\/h3>\n

llama.cpp<\/code>, which forms the core of the inference engine, has been updated to v0.4.1. This may improve support for the latest models and the stability of inference processing.<\/p>\n

WSL2 and GPU Passthrough Fixes<\/h3>\n

Multiple fixes related to detection, image selection, and GPU device passthrough in WSL2 environments have been implemented. Operation stability is improved for users utilizing GPU acceleration via the WSL2 backend using Docker Desktop or Podman Desktop on Windows.<\/p>\n

Prevention of File Mixing in Snapshots<\/h3>\n

An issue where GGUF format and safetensors format files could become mixed during snapshot creation has been fixed. This prevents potential inconsistencies that could occur when saving or loading models.<\/p>\n

Full Support for Toolbox Environments<\/h3>\n

Running inside a toolbox using containers is now fully supported. This is suited for users who want to run RamaLama within more isolated environments or specific development container environments.<\/p>\n

Podman 6 Support on macOS<\/h3>\n

When using Podman 6 on macOS, a fix was made so that calls to podman-machine-list<\/code> function correctly. This impacts users utilizing Podman in macOS environments.<\/p>\n

Update to huggingface-hub Dependencies<\/h3>\n

Dependencies on huggingface-hub<\/code> have been updated to support the latest version (~=1.32.0). This improves the stability of fetching models from Hugging Face.<\/p>\n

Supported Models and Hardware<\/h2>\n

With this update, the update of the inference engine llama.cpp<\/code> to v0.4.1 potentially broadens the range of supported model architectures and quantization formats. Furthermore, enabling Vulkan in NVIDIA GPU environments and fixing GPU device passthrough in WSL2 allows for more stable inference leveraging NVIDIA GPUs from container environments on Windows.<\/p>\n

How to Update<\/h2>\n

Updates can be performed using the following commands depending on your environment.<\/p>\n

When Using PyPI<\/h3>\n
pip install --upgrade ramalama\n<\/code><\/pre>\n
When Using Fedora<\/h3>\n
sudo dnf upgrade ramalama\n<\/code><\/pre>\n
When Using the Install Script<\/h3>\n
curl -fsSL https:\/\/ramalama.ai\/install.sh | bash\n<\/code><\/pre>\n
If you are using the .pkg<\/code> installer on macOS, please download the latest installer from Releases<\/a> and reinstall.<\/p>\n
<\/p>\n
Releases Since Our Last Article<\/h2>\n
Compiled by Local Model Watch from the project’s GitHub releases: the versions between this release and the last one we covered, which did not get separate articles.<\/em> Full history: release tracker<\/a>.<\/em><\/p>\n
\n\n\n\n\n\n\n\n\n\n
Version<\/th>\nReleased<\/th>\nRelease notes<\/th>\n<\/tr>\n<\/thead>\n
v0.24.0<\/td>\n2026-08-22<\/td>\nGitHub<\/a><\/td>\n<\/tr>\n
v0.23.0<\/td>\n2026-06-25<\/td>\nGitHub<\/a><\/td>\n<\/tr>\n
v0.22.0<\/td>\n2026-06-06<\/td>\nGitHub<\/a><\/td>\n<\/tr>\n
v0.21.0<\/td>\n2026-05-14<\/td>\nGitHub<\/a><\/td>\n<\/tr>\n
v0.20.0<\/td>\n2026-04-28<\/td>\nGitHub<\/a><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n
<\/p>\n
<\/p>\n
What to Read Next<\/h2>\n