The July 27, 2026 QVAC SDK 0.16 release added image, video and robotics features, while Tether Data also open-sourced VisionPsy-Nano, a roughly 460 million-parameter vision-language model for local and edge deployment.
Tether has expanded its QVAC artificial intelligence stack with the July 27, 2026 release of QVAC SDK 0.16 and the open-source launch of VisionPsy-Nano, a roughly 460 million-parameter vision-language model aimed at on-device and edge deployment. The SDK update adds image generation through Ideogram 4, synchronized video with audio via LTX-2, local execution of NVIDIA’s GR00T robotics model, and improvements to an OpenAI-compatible server for coding agent integration. Tether said VisionPsy-Nano achieved the highest normalized score, 62.3, among evaluated sub-500M on-device models and outperformed rivals in 16 of 17 benchmarks. Both the SDK and VisionPsy-Nano are distributed under the Apache 2.0 license. The broader QVAC toolkit runs across multiple operating systems through Node.js, supports multimodal AI spanning text, vision, speech and robotics, and uses the Holepunch stack for peer-to-peer capabilities. CEO Paolo Ardoino has framed the initiative as part of a "Stable Intelligence Era," linking local AI execution to themes of privacy, control and financial sovereignty associated with Tether’s broader ecosystem. Tether has no native token tied to QVAC, and there is no QVAC coin.