Astra
About

About Astra

The name Astra comes from the Latin word for stars. We believe intelligence shouldn't be locked away in cloud datacenters — it should be like starlight: within reach, for everyone.

The past few years have seen an explosion of open-source LLMs — Llama, Qwen, Gemma, DeepSeek, Phi. They're powerful, yet small enough to fit in a pocket. And llama.cpp — that open-source miracle — lets a C++ inference engine run smoothly on a phone CPU.

What Astra does is wrap all of this into a simple app: install, chat, pull more models — no need to understand GGUF, quantization, or the command line. And most importantly, it all stays local. Your conversations belong only to you.

Principles

What we stand for

Privacy first

Local inference is the default — not a premium feature.

Fully open-source

Standing on llama.cpp and the open model ecosystem.

Users first

No ads, no tracking, no forced accounts.

Always evolving

New models, new platforms, new capabilities — always shipping.

Tech Stack

Tech stack

llama.cpp
An open-source C/C++ inference engine by Georgi Gerganov, supporting GGUF and quantized inference.
Metal · Vulkan · OpenCL
Optimized for iOS and Android GPU acceleration.
Hugging Face Hub API
The world's largest open model hub, integrated end-to-end.
ModelScope
Alibaba DAMO's model community — fast access from mainland China.
Join us

Onward, to the stars

Astra is a long-term project. If you believe in local-first and open-source, walk with us.