Tencent Hunyuan open-sources Hy Memory 2.0 and memory writer model
Tencent Hunyuan released Hy Memory 2.0, an agent memory framework, and open-sourced three layers: the framework SDK, the local MemHub memory app, and a dedicated memory-writer model, Hy-Memory-30B-A3B (~3B active parameters, Apache-2.0). According to Tencent and its accompanying paper, the model scores a 55.49% macro average across four harnesses (Mem0, MemOS, Hy-Memory and held-out EverMemOS) on fixed 20% benchmark subsamples, edging past Claude Opus 5 at 54.62%, DeepSeek V4.1 Flash at 53.64% and Qwen3-30B-A3B at 44.37%. The framework covers chat, working and code memory — Tencent calls it the first to propose working and code memory — and is already used in internal finance, news, knowledge-management and office-agent businesses, while MemHub plugs into about 10 AI tools including WorkBuddy, Cursor, Codex, Qwen Code and DeepSeek Harness. Why it matters: Memory writing is becoming a critical piece of long-horizon agent tasks, and a small, transferable 3B writer model offers a reproducible data point on whether the memory layer deserves its own trained model. What do you think?
Open the headline: source, AI briefing and more →Comments
…