Gradient open-sources Parallax to promote the implementation of native AI applications

Foresight News reported that the distributed AI laboratory Gradient has announced the open-sourcing of Parallax, an operating system designed for local AI applications. This system supports cross-platform and cross-region deployment of open-source large models on heterogeneous devices such as Mac and Windows, allowing users to have full control over models, data, and AI memory. Parallax is equipped with built-in network-aware sharding and dynamic task routing mechanisms, enabling intelligent scheduling based on inference workloads and seamless switching between single-machine, multi-device, and wide-area cluster modes.

Currently, Parallax is compatible with more than 40 open-source large models, including Qwen3, Kimi K2, DeepSeek R1, and gpt-oss. Developers can deploy locally to build and run various AI applications such as programming assistants, personal agents, and multimodal generation in a fully autonomous manner, keeping all sensitive data and control permissions locally.