Tether, issuer of the stablecoin USDT, has launched an AI training framework designed to fine-tune large language models on consumer hardware, including smartphones and non‑Nvidia GPUs. The system is part of Tether’s QVAC platform and leverages Microsoft’s BitNet architecture alongside LoRA techniques to cut memory and compute needs, aiming to lower costs and hardware barriers for model development.
The cross‑platform framework supports training and inference across AMD, Intel and Apple Silicon chips, plus mobile GPUs from Qualcomm and Apple. Tether says its engineers fine‑tuned models of up to 1 billion parameters on smartphones in under two hours, with smaller models taking minutes, and support extending to models as large as 13 billion parameters on mobile devices.
Built on BitNet’s 1‑bit model design, the framework can reduce VRAM requirements by as much as 77.8% compared with equivalent 16‑bit models, enabling larger models to run on limited hardware. It also enables LoRA fine‑tuning for 1‑bit models on non‑Nvidia hardware, expanding beyond the GPUs conventionally used for AI training.
Tether reports inference performance improvements too, with mobile GPUs running BitNet models multiple times faster than CPUs. The company highlights on‑device training and federated learning as potential use cases, where models are updated across distributed devices without centralizing user data—potentially decreasing dependence on cloud infrastructure.
Tether’s move into AI infrastructure follows a broader trend of crypto firms expanding into compute and machine learning, particularly as Bitcoin mining companies pivot toward AI and high‑performance computing (HPC). Recent industry activity includes Google taking a 5.4% stake in Cipher Mining as part of a long‑term AI data center deal, Bitcoin miner IREN planning multibillion‑dollar capital raises for AI infrastructure, HIVE Digital Technologies reporting record revenue driven by AI and HPC, and Core Scientific securing a large credit facility for data center development.
The sector is also seeing growth in AI agents—autonomous programs that transact and interact with services. Coinbase has introduced wallet infrastructure for AI agents to perform on‑chain transactions; Alchemy launched a system allowing agents to access blockchain data services using USDC on Base; and Pantera and Franklin Templeton joined Arena for testing enterprise AI agents. World, the identity network co‑founded by OpenAI’s Sam Altman, launched AgentKit to let AI agents verify links to unique humans via World ID while making payments with the x402 micropayments protocol.
Cointelegraph states this article follows its Editorial Policy and aims to provide accurate, timely information; readers are encouraged to verify independently. Read the policy at https://cointelegraph.com/editorial-policy
