QVAC Workbench

QVAC Workbench is a desktop application developed with React and the QVAC SDK that allows loading, managing, and running artificial intelligence language models directly on the local computer. The application runs on the Pear runtime, providing a secure and distributed computing environment. QVAC Workbench puts powerful AI models on the user's device, offering experimentation, customization, and exploration of artificial intelligence capabilities with total privacy and without relying on the cloud. The platform allows querying locally stored documents, processing the information without sharing it with third parties, ensuring absolute control over the data.

★★★★★

5.0(1 ratings)

• Local model loading. Allows loading language models in QVAC format from the computer's internal storage. Models are verified using a checksum to guarantee their integrity before execution.
• Offline execution. All inference processes are carried out entirely on the local hardware. No communication with external servers is established during the use of the models, ensuring the privacy of the queries.
• Model manager. Provides an interface to organize installed models, display their version, size, configuration parameters, and memory requirements. It allows enabling or disabling models according to the user's needs.
• Document queries. The user can upload documents in PDF, TXT, or Markdown formats and ask questions about their content. Text processing is done locally without sending the documents to cloud services.
• Parameter customization. Offers controls to adjust the temperature, token limit, maximum context, and other hyperparameters of the language models. Changes are applied in real-time during execution.
• Conversation history. Maintains a complete record of interactions with the models. Conversations are stored in encrypted format on the local disk and can be exported for later analysis.
• Compute delegation mode. Allows transferring the processing load to more powerful devices within the same local network controlled by the user. Communication between devices uses point-to-point encryption.
• Custom model import. Supports importing user-trained models as long as they comply with the QVAC SDK specification. The system validates compatibility before integrating the model into the work environment.
• Performance monitor. Displays real-time CPU, RAM, and GPU usage (if available) during model execution. Includes latency and token generation speed graphs per second.
• Isolated workspaces. Each project can be configured with its own set of models and documents in an independent execution environment. Workspaces are separated using sandboxing to prevent interference.
• Result export. Responses generated by the models can be exported in JSON, CSV, or plain text formats. It is also possible to generate reports that include the questions asked and the answers obtained along with execution metadata.
• Component updates. The system periodically checks for new versions of the Pear runtime and the underlying SDK. Updates are downloaded and installed in a controlled manner after user confirmation.

The development of QVAC Workbench began in 2025 by Tether's engineering team, specifically within the division focused on decentralized infrastructure. The application is primarily written in JavaScript using the React framework for the user interface, while the low-level modules that interact with the AI models are implemented in Rust to ensure efficiency and safety in memory management.

Alternatives to QVAC Workbench:

Ollie IDE

Ollie is a sovereign creative suite designed for developers, writers, and creators who want the power of AI without sacrificing privacy or paying monthly subscriptions.

Price: $19   Size: 180 MB   Version: 2026.2.28   OS: Windows, Linux, MacOS

LM Studio

LM Studio is a desktop application that allows you to run large language models (LLMs) such as GPT-oss, Qwen3, Gemma3, and DeepSeek directly on a personal computer.

Price: Free   Size: 568 MB   Version: 0.4.8-1   OS: Windows, Linux, MacOS