The tweet was deleted by the author.
But we saved everything 🙂.
Paolo Ardoino, Chief Technology Officer at Tether, showcases the impressive speed of QVAC running local inference on a mobile device. The demonstration utilized llama.cpp combined with the advanced machine learning model LLAMA 3.2, featuring 1 billion parameters.
QVAC, described as a versatile inference and fine-tuning runtime, is designed to adapt seamlessly across a variety of devices including smartphones, laptops, and servers. The innovation reflects Tether's forward-thinking approach to technological advancement in the crypto space, highlighting the potential of deploying complex models on accessible hardware for everyday use.
Tether’s efforts to broaden the utility of advanced AI models reiterate the company’s broader commitment to technical innovation, a strategy reflected in Ardoino’s earlier introduction of PearPass, promising local and private data sync capabilities. These advancements take place against a backdrop of financial rigor, underscored by Tether’s stablecoin leadership and the confirmation that USDT reserves hold $127 billion in U.S. Treasury, reinforcing confidence in both the firm’s technological and fiscal stewardship.