Hy-Memory-30B-A3B is an open-weight model from Tencent Hunyuan that does one job for AI agents: deciding what to write into their long-term memory.
In plain words
Agents that remember earlier sessions keep a memory bank, and a general language model usually decides what goes in. Hy-Memory-30B-A3B is a specialist for that step: it chooses what to keep, how exactly to record it, and how new information should change what is stored. Only about 3 billion of its parameters are active at a time. It takes the place of the general model inside memory tools such as Mem0 and MemOS. The weights are on Hugging Face under the Apache 2.0 licence.
A quick example
Imagine a study helper app you chat with every week. After one chat, a memory model notes that you switched from biology to chemistry, and it updates the old note instead of keeping two that disagree.
Where you'll meet it
- Chat assistants that need to remember a user across many conversations.
- Coding agents that keep track of how a project changes over time.
Related
Sources
- DocsHy-Memory-30B-A3B model card, Tencent Hunyuan (Hugging Face) · read 16 Oct 2026
- DocsHy-Memory, Tencent Hunyuan (GitHub) · read 16 Oct 2026
Checked by WhatIsAI against the sources above · 16 Oct 2026. A full five-minute explainer with a diagram may follow. Spotted an error? Report it.