Tether, the company best known for issuing the USDT stablecoin, has officially launched the QVAC SDK, an open-source software development kit designed to enable AI applications to run fully on-device, without reliance on cloud servers. The company positions the move as a foundational step toward what it calls the "Stable Intelligence Era," pushing its brand into decentralized infrastructure.
The QVAC SDK is built on a customized branch of the llama.cpp framework, dubbed QVAC Fabric. It supports core AI capabilities including text generation, speech processing (via whisper.cpp and Parakeet), visual recognition, and on-device translation (via Bergamot), all accessible through a unified API. A key feature is its cross-platform compatibility, allowing a single codebase to run identically on iOS, Android, Windows, macOS, and Linux, significantly reducing development complexity.
Privacy and decentralization are central to the offering. By processing data locally, the SDK aims to eliminate the need to send sensitive user information to remote servers. The system can function without an internet connection and is resilient to centralized server outages. For peer-to-peer model distribution and delegated inference, the SDK utilizes the Holepunch protocol stack, allowing devices within a network to share computational workloads and updates.
In a statement, Tether CEO Paolo Ardoino criticized centralized AI as a "dead end," citing issues with latency, single points of failure, and concentrated control. He stated QVAC is built "for the world of the next billions of years."
Looking ahead, Tether has outlined an ambitious roadmap for the QVAC ecosystem. The company plans to add decentralized training and fine-tuning capabilities on top of the current inference tooling. Furthermore, Tether announced it will allocate resources to develop specialized toolkits for robotics and brain-computer interface applications, aiming to evolve the project into a full-stack, distributed AI environment.
