Tether has launched QVAC SDK, enabling local AI deployment across devices without cloud reliance.
Tether has launched the QVAC SDK, a toolkit aimed at facilitating the development and deployment of artificial intelligence on various devices, ranging from smartphones to servers.
Tether's QVAC SDK serves as a universal AI building block, designed for running, training, and evolving intelligence on any platform.
Learn more: https://t.co/n7mkjkD8cz
— Tether (@tether) April 9, 2026
This open-source toolkit enables users to create and retrain AI models without relying on cloud services, allowing for a seamless experience across iOS, Android, Windows, macOS, and Linux through a singular codebase.
Shifting Towards Decentralization
Having introduced the QVAC platform for decentralized AI in May 2025, Tether emphasizes the limitations of current centralized AI systems, which face challenges like latency and lack of scalability. The QVAC approach enables local processing, meaning that user data remains on the device rather than being sent to external servers.
This method aims to enhance:
- Privacy protection;
- System responsiveness;
- Reliability during connectivity issues.
“Applications using QVAC remain operational even in low connectivity situations, providing practical AI solutions in real-world applications,” explained the developers.
Core Technical Features
The SDK is built on the QVAC Fabric, employing a modified version of the llama.cpp engine for local model execution. It is compatible with the Llama ecosystem for various AI tasks like text generation and translation.
Additionally, the QVAC SDK encompasses several autonomous features like Parakeet for speech recognition and Bergamot for translation, all accessible via a unified API, simplifying the integration process for developers.
With built-in peer-to-peer functionalities based on the Holepunch stack, the SDK allows for model distribution without central servers, promoting a distributed network for both the usage and training of AI.
Tether envisions the QVAC SDK as a cornerstone for next-generation applications, ranging from personal AI assistants to autonomous agents within Web3 frameworks.
“As we approach a future where billions of people coexist with numerous autonomous machines and trillions of AI agents, the reliance on centralized servers will become unfeasible,” remarked Tether CEO Paulo Ardoino.
Earlier this year, the QVAC team also released an open dataset to support AI training initiatives.
