A new open-source LLM gateway unifies self-hosted, frontier, and open models behind one API, adds sub-millisecond overhead, takes no token markup, and can use opt-in traffic to route smarter or train a custom model.
After a CEO laid off engineers to cut costs in favor of AI tooling, those same engineers shipped OpenExecutive — an open-source agent designed to handle executive decision-making. Turnabout, apparently, is fair play.
Meta's Muse Glimmer is an open-weight model that signals where the company is heading with personalized AI — and sharpens the contrast between AI you can run yourself versus AI locked behind a platform.
llama.cpp lets you run large language models on your own hardware—CPU or GPU—without cloud dependencies. Here's why it became the backbone of the local AI movement and how to put it to work.
A new Qwen release surfaced on Hugging Face with an FP8 checkpoint and a mixture-of-experts design. Here's how to read the model card details and decide whether it fits your stack.
Meta shipped Glimmer, a downloadable open-weight model, while keeping its stronger Muse Spark behind private APIs. The split reveals how 'open AI' works in practice — and what builders actually get.
Alibaba's Qwen team publishes a 27B-parameter model in FP8 format, pushing open-weight models toward lower memory footprints and faster inference on modest hardware.
Meta released Glimmer, a downloadable open-weight model, while keeping its more capable Muse Spark locked behind proprietary APIs — a split that defines what 'open AI' really delivers in practice.
MacPaw is rebuilding its Eney AI assistant to run entirely on-device, powered by Liquid AI's models — and plans to expose that inference capability to developers building on its platform.
A new open-source project claims to run an 80-billion-parameter Qwen model in just 4.3 GB of RAM on a Mac, and a 35B model on an iPhone. Here's what that means for local inference.
OpenAI has published Codex Security as an open-source repository, surfacing the threat models, security principles, and guidelines that govern how their AI coding agent handles untrusted code and environments.
Thinking Machines has published Inkling as an open-weights model, giving builders direct access to run, fine-tune, and deploy it on their own terms rather than through a locked API.
