Tencent Hy Memory 2.0 goes open source: 10 AI tools share memory, even rolled-back code can be traced

Beating AI News Flash: Tencent Hunyuan has upgraded Hy Memory to 2.0, open-sourcing a memory framework, a dedicated model, and the MemHub application all at once.

The 1.0 version launched in May primarily handled chat memory. It could track changes in the same matter over time, detect and resolve contradictions, and summarize unspoken user habits from multiple conversations. When users changed preferences, the system could retain old records while remembering the latest decisions.

2.0 adds working memory and code memory. It records how far a task has progressed, what approaches were adopted, and what the final results were. When encountering similar tasks, it also checks whether past experience is applicable. Code memory goes further, integrating Git records to track the modification history of files and functions. Even if a piece of code was later rolled back, it can retrieve the changes made at the time and identify what methods were previously tried.

The accompanying MemHub connects memory across different AI tools. 10 tools including Cursor, Codex, and Claude Code can share a single local memory. With user authorization, MemHub can import historical conversations and continuously sync new records. For example, if project specifications are established in Cursor and development continues in Codex, they can be directly retrieved without re-explaining.

Tencent also trained a dedicated model, Hy-Memory-30B-A3B, with 30 billion total parameters and only about 3 billion activated per inference. It specializes in filtering, organizing, and updating memories, and can also integrate with existing frameworks such as Mem0 and MemOS. Tencent tested four memory systems, with an average score of 55.49, slightly higher than Opus 5's 54.62.