Tether Data Launches QVAC Fabric LLM to Enable LLM Training on Consumer Devices

As reported by HashNews, Tether Data has announced the launch of QVAC Fabric LLM, a new large language model inference runtime and fine-tuning framework. The technology allows users to run, train, and customize large language models directly on consumer-grade GPUs, laptops, and smartphones, eliminating the need for high-end cloud servers or NVIDIA systems. QVAC Fabric LLM supports models such as Llama3, Qwen3, and Gemma3, and expands the llama.cpp ecosystem. Training is supported on GPUs from AMD, Intel, NVIDIA, Apple, and mobile chips. The framework is open-sourced under the Apache 2.0 license, with pre-built binaries and ready-to-use adapters available on Hugging Face, enabling developers to fine-tune models with just a few commands.