Tether Launches QVAC SDK and Developer Grants Program for Local-First AI — Open-Source Cross-Platform Framework Runs LLMs, Speech, and RAG Offline on Any Device

GitHub·medium signal

Tether launched a developer grants program on May 11 to fund projects building on QVAC, its open-source SDK for on-device AI agents that work offline and sync peer-to-peer. QVAC integrates llama.cpp (via QVAC Fabric), whisper.cpp and Parakeet for speech-to-text, Bergamot for on-device translation, and Holepunch for peer-to-peer model distribution — all running across Linux, macOS, Windows, Android, and iOS. The grants program signals a serious attempt to build a developer ecosystem around local-first AI as an alternative to cloud-dependent inference.