Tencent Hunyuan has officially released Hy-Memory, a memory plugin designed for long-term collaborative AI agents. It uses a 6-layer memory framework and dual System1/System2 processing to enable agents to remember durably and efficiently, reducing memory count by over 70% and token usage by 35% on ultra-long contexts. This aims to move agents beyond single-session context, allowing them to build a persistent, evolving understanding of user preferences and intentions.
Tencent Hunyuan has officially released Hy-Memory, a memory plugin built to provide long-term memory for AI agents (autonomous systems that plan and act). It features a 6-layer memory framework, a System1/System2 dual processing system, and a three-layer evolutionary chain. This architecture allows agents to retain and process information more durably, accurately, and efficiently.
- Memory Reduction
- 70%+ fewer memories
- Information Density
- 45%+ higher per memory
- Token Usage Reduction
- 35% less on ultra-long contexts
- LongMemEval Score
- 85.20%
- PersonaMem Score
- 76.91%
- Write Time (vs Graphiti)
- 8x faster
The release addresses common challenges in agent memory, such as fragmentation and inefficient context management, which often lead to high token usage (the units of text processed by an AI model). By structuring and evolving memories, Hy-Memory aims to improve an agent's understanding of user preferences and intentions over time. This aligns with efforts by companies like Oracle AI Agent Memory, Anthropic Native Memory, and MiniMax M3 to give agents persistent learning capabilities.
Hy-Memory offers three tiers: Lite for vector retrieval, Pro for synchronous memory processing, and Ultra for the full System1 and System2 kernel. This tiered design targets the memory bloat and token costs that limit long-running agents.
