- Tether unveiled a game-changing upgrade for the QVAC Fabric, expanding its AI fine-tuning capabilities straight to regular consumer devices, such as smartphones.
- The platform eliminates the need for enterprise-grade GPUs and centralized cloud servers in AI agent development and maintenance, thereby reducing exposure of users’ data to tech institutions.
Tether continues building on its decentralized AI (artificial intelligence) on-device infrastructure. Barely a year after the stablecoin giant unveiled the QVAC (QuantumVerse Automatic Computer) in May 2025, it has expanded the platform’s capabilities for more practical applications.
Tether’s QVAC Fabric Upgrade
On Tuesday, Tether announced an upgrade to the QVAC Fabric, integrating new features to turn it into the world’s first cross-platform 1-bit LLM (Large Language Model) LoRA (Low-Rank Adaptation) fine-tuning framework. It extends Microsoft’s ultra-efficient BitNet architecture to allow AI configuration, training, or inference on everyday hardware, including smartphones, laptops, and consumer GPUs (Graphics Processing Units).
The highly efficient, low-resource customization features of the QVAC Fabric bypass the need for expensive enterprise-grade Nvidia systems or cloud infrastructure in training and maintaining AI models. It integrates AI development and management directly into heterogeneous consumer GPUs, including Intel, AMD, Apple Silicon M chips, and others. As a result, it expands the capability of regular consumer devices, such as smartphones, for various AI use cases.
The breakthrough is compatible with widely available and popular consumer devices. These include iPhones, Samsung Galaxy phones, Google Pixel phones, and desktops and laptops running Vulkan- and Metal-based backends.
Why It Matters
QVAC Fabric utilizes Apache 2.0 on Hugging Face, enabling up to 90% memory reduction and 2x to 11x faster inference even without high-capacity GPUs. For crypto use cases, the infrastructure unlocks new possibilities for decentralized finance (DeFi) and self-sovereign data management.
For example, an AI model that typically requires 40 gigabytes of RAM can now be streamlined to run on hardware with less than 4 GB of RAM, equivalent to the specs of mid-range smartphones.
The platform paves the way for users to personalize their AI agents without necessarily connecting to centralized servers. By making the AI agents live entirely on portable hardware, it mitigates risks related to data privacy and centralized censorship. It particularly prevents the Big Tech cloud providers from harvesting their information, particularly personal keys that underpin the security of their digital assets.
Paolo Ardoino, CEO of Tether, believes AI will be the primary driving force of the future. He described it as an element that would bring greater stability, cohesion, and empowerment to a society.
Furthermore, Ardoino highlighted that reliance on centralization in training LLMs could only lead to stagnation in innovation and make the AI ecosystem more fragile. By making the technology more accessible to regular devices, it enables the democratization of intelligence on a global scale, aligning with Tether’s Stable Intelligence roadmap with the QVAC.
