RightSignal · Friday, November 13

Open sourceggml-org / Hugging Face

llama.cpp 1.0

llama.cpp, a popular tool for running AI models on your own computer, has reached version 1.0. Llama desktop apps built on it were announced at the same time.

What’s new: For people who run AI locally. The team says it will keep covering every platform and model type, and the new apps aim to make local AI a few clicks to set up.

Open sourceHugging Face

Transformers 5.23.0

A new release of the widely used Python library for AI models. It adds support for models including Inkling 2 Small, LFM3, LFM3-VL and InternLM3, plus a new way to export models.

What’s new: For developers building with open models. Exports now return one object you can run, share on the Hub and load back. Several changes break older code.

GitHub release

Open sourceRust project

Rust 1.100.0

A new version of the Rust programming language. It makes the never type and custom memory allocators stable features.

What’s new: For Rust programmers. Custom allocators let collections like Vec and Box use your own memory allocator. wasm32-wasip3 also becomes a Tier 2 platform.

GitHub release

New modelThinking Machines Lab

Inkling 2 Small

An open-weight model that takes text and images and writes text. It uses a mixture-of-experts design with 594B total and 17B active parameters.

What’s new: For developers who want a large open multimodal model. The weights are now on Hugging Face, and Transformers supports it.

221 likes on Hugging Face

Hugging Face

New featureGitHub

GitHub Copilot app: Enterprise Server support

The GitHub Copilot app can now connect to GitHub Enterprise Server accounts, so you can work with their repositories, issues and pull requests.

What’s new: For companies that host GitHub on their own servers. The feature was an opt-in preview and is now generally available from app version 1.2.0. Cloud sessions are not supported.

GitHub Changelog

That’s every launch.

Next: 5 engineering blogs from the same day.