• Tether has expanded the QVAC SDK.
• AI can now create video and decode neural signals locally.
• In addition, Tether's new local AI toolkit has gained robotics support.

Tether's research division, Data AI, has unveiled the QVAC SDK 0.13.0 update. The release adds support for local video generation, brain-computer interfaces (BCI) and robotics models, and simplifies the creation of desktop AI applications. The release of the new version was announced by Tether CEO Paolo Ardoino.

«Tether AI progress. Two weeks ago we released TurboQuant support in QVAC, and today we have improved the entire stack for developers», he said.

According to the team, the QVAC SDK 0.13.0 makes it possible to integrate local AI models directly into development environments and agentic programming systems. 

The platform also gained support for new OpenAI-compatible APIs and a tool for creating desktop applications for macOS, Windows and Linux with a single command.

The update is part of Tether's strategy to develop local AI infrastructure without dependence on cloud services.

A guide to antifragility — how and why did Tether transform from a niche company into a multi-industry conglomerate? 22.05.2026 Read

Local video, robotics and BCI

One of the key innovations was support for the Wan 2.1 model for image-to-video generation in local mode. Developers can create video clips without transferring data to third-party services.

In addition, QVAC added support for the pi0.5 model, which the company calls a «brain for robots». It makes it possible to run multimodal robot control systems on a single consumer graphics card without loss of accuracy.

The release also introduced a new BCI module for brain-computer interfaces. The solution is based on whisper.cpp and allows recorded neural signals to be converted into text locally without transferring the information to third parties.

Among other improvements:

• accelerated GPU transcription of Whisper on Android;
• improved performance of Parakeet voice models on mobile devices;
• full GPU acceleration for the Supertonic TTS speech synthesis system via Metal, Vulkan and OpenCL;
• more detailed statistics on model performance and graphics memory usage.

The development of the QVAC ecosystem

The release continues the series of Tether AI products presented over recent months.

Tether AI: why did the issuer of the USDT stablecoin begin its expansion into the AI market? 05.05.2025 Read

In December 2025, the company introduced the QVAC-fabric-llm system for local fine-tuning of large language models, which makes it possible to run LoRA training on any devices — from smartphones to servers.

Later, Tether launched the QVAC Health platform for analyzing personal health data and presented the QVAC Genesis II dataset, which after expansion covered 148 billion tokens across 19 educational domains.

Source: Incrypted