# App developers turn to QVAC for AI tooling without the cloud

> Source: <https://www.networkworld.com/article/4250425/app-developers-turn-to-qvac-for-ai-tooling-without-the-cloud.html>
> Published: 2026-11-18 21:14:44+00:00

**Key takeaway:** Tether's QVAC SDK pairs a llama.cpp-based Fabric engine with the TurboVec vector index, which cuts embedding memory use up to 8x at 4-bit quantization so RAG can run on-device instead of in the cloud.

In most instances today, adding AI to an app means sending user data to a third-party API, paying per request, and hoping the connection holds.

Tether’s [QVAC](https://qvac.tether.io/) is built on the opposite premise. It runs AI models directly on the user’s device. Data never leaves the hardware, there are no per-request costs, no rate limits, and no need for an internet connection once a model is downloaded.  

**QVAC SDK: the next milestone in cloud-free AI** 

While open-source tools and standards like llama.cpp, GGML, and ONNX have made local AI much more accessible in recent years, creating new possibilities for consumer devices, the edge AI ecosystem remains fragmented.

Against this backdrop, in 2026, Tether released a software development kit (SDK) that pulls together the best open tooling from across different domains of machine learning.

The centerpiece of Tether’s [QVAC](https://qvac.tether.io/products/sdk) [SDK](https://qvac.tether.io/products/sdk) is [Fabric](https://qvac.tether.io/products/fabric). Built on llama.cpp, it expands the engine to enable on-device fine tuning and additional AI capabilities—opening access to memory-intensive workloads previously only possible with expensive hardware or cloud services.   

Building on GGML’s modular framework, QVAC Fabric is primarily oriented toward three backends: Vulkan, which covers most consumer-grade desktop and mobile chips; Metal and Core ML for Apple Silicon; and CUDA for higher-end NVIDIA GPU setups.

The QVAC SDK empowers developers with a unified interface for building multiplatform [AI](https://techcrunch.com/sponsor/tether/on-device-ai-is-ready-to-challenge-cloud-dominance/) [apps](https://techcrunch.com/sponsor/tether/on-device-ai-is-ready-to-challenge-cloud-dominance/) [that](https://techcrunch.com/sponsor/tether/on-device-ai-is-ready-to-challenge-cloud-dominance/) [run](https://techcrunch.com/sponsor/tether/on-device-ai-is-ready-to-challenge-cloud-dominance/) [on-device](https://techcrunch.com/sponsor/tether/on-device-ai-is-ready-to-challenge-cloud-dominance/). But although significant progress has been made in bringing inference to consumer devices, text generation alone is only one part of the pipeline that makes modern AI useful.   

**Bringing search to edge AI**  

RAG (retrieval-augmented generation) is what equips AI to import and process contextual information to enhance its output. Instead of language models relying solely on their training data to answer questions, RAG lets them answer using a specific set of documents.

Implementing RAG into an AI app incurs significant memory overhead versus running a model alone. Alongside the core language model’s inference, RAG requires a second embedding model and a specialized database system known as a vector index.

These vector indexes can be huge, often much larger than the plain text files they represent and easily surpassing the RAM capacity of consumer hardware. As a result, most RAG operations are offloaded to the cloud, along with the user’s personal information. But vector indexing doesn’t have to mean sacrificing privacy or getting locked in to pay-per-token subscriptions.

In September 2026, the QVAC SDK [integrated](https://qvac.tether.io/blog/turbovec-local-vector-search) [TurboVec](https://qvac.tether.io/blog/turbovec-local-vector-search). The open-source vector index shrinks embedding memory usage by up to 8 times (4-bit quantized). This lowers hardware requirements and significantly increases the speed of search and retrieval, paving the way for RAG to be integrated into offline, cloud-free apps that can access local databases without leaking personal information. 

**Lightweight AI models**  

Running AI locally only works if it fits the hardware people actually own. Frontier-scale systems typically need data-center GPUs and hundreds of gigabytes of memory. That’s why the QVAC SDK integrates quantized versions of common open-source models, while Tether has also developed its own lightweight models designed to run on a phone or personal computer, and distributed peer-to-peer via its registry.

QVAC Psy is a family of specialized low-weight models that run on the QVAC SDK. Using specially curated training data, each model is trained to perform specific tasks at the same or a higher level than much larger AI platforms. For example, despite weighing in at just 800 million parameters, the smallest variant of Tether’s [translation](https://tether.io/news/tether-releases-open-source-ai-translation-models-for-african-and-european-languages/) [model](https://tether.io/news/tether-releases-open-source-ai-translation-models-for-african-and-european-languages/) [for](https://tether.io/news/tether-releases-open-source-ai-translation-models-for-african-and-european-languages/) [African](https://tether.io/news/tether-releases-open-source-ai-translation-models-for-african-and-european-languages/) [languages](https://tether.io/news/tether-releases-open-source-ai-translation-models-for-african-and-european-languages/) outperformed Google’s 27 billion parameter TranslateGemma on key benchmarks.  

By pairing these compact models with other memory-saving features packed into the QVAC SDK, Tether is proving that local AI doesn’t mean compromising on performance. And once apps can deliver real-world utility without the need for massive data centers, users get to enjoy unlimited, secure intelligence for free, without an internet connection.

**Why you should consider local AI**

Tether’s tooling has the potential to transform how AI is built and used, with benefits for developers and end users alike.

Without the burden of recurring API costs or expensive cloud hosting, independent creators and smaller studios can implement AI features that previously required enterprise-scale budgets. Meanwhile, processing sensitive information directly on local hardware addresses privacy concerns and can help to ease compliance with regulations like [GDPR](https://www.computerweekly.com/opinion/GDPR-wasnt-designed-for-AI-and-thats-a-security-problem) and HIPAA, opening the door for secure AI apps in the healthcare, legal and financial sectors.  

If you’re ready to eliminate ongoing API fees and take full control of your application’s data, download the [QVAC](https://qvac.tether.io/products/sdk) [SDK](https://qvac.tether.io/products/sdk) and join the growing community of builders rejecting the cloud.
