Tencent Hunyuan Releases Hy Memory 2.0: Fully Open Source Across Model, Framework, and Application Layers, Ranked First in Open Source Memory Framework Evaluation

Beating AI News Flash: Tencent Hunyuan has announced the upgrade of its intelligent agent memory framework, Hy Memory, to version 2.0, along with the open-sourcing of the dedicated model Hy-Memory-30A3B, the memory framework, and the MemHub application. The three-layer capabilities can be used independently or combined into a complete memory solution. The framework handles the storage, updating, and evolution of three types of memories: chat, work, and code. It ranks first among open-source memory frameworks across four benchmarks. The dedicated model scored 55.49 in evaluations on multiple memory systems such as Hy Memory, Mem0, and MemOS, surpassing Opus5's score of 54.62. Hy Memory 2.0 upgrades chat memory from fragmented facts to complete event contexts and introduces work memory and code memory, enabling the judgment of experience applicability and precise assistance for long-term tasks. MemHub, as a local memory management application, supports CLI and Mac platforms and can connect with 10 AI tools such as WorkBuddy, Cursor, and Codex, enabling historical memory import and real-time synchronization across tools. This capability has already been implemented in Tencent's Stock Selector, iWiki, Tencent News, and other services: the multi-hop recall rate for Stock Selector increased from 50% to 90%, with a comprehensive score improvement of 23.8%; iWiki's accuracy when answering questions using only memory improved by 48.6% compared to providing all original conversations; Tencent News achieved an average score of 91.68 in long-term preference evaluations.