Tencent Hunyuan open-sources Hy Memory 2.0 agent memory framework
10/16/2026 — 10/17, 03:17·2 sources (all)·2 reports (all)
Key takeaways
- Tencent Hunyuan has released and open-sourced Hy Memory 2.0, an agent memory framework with three layers: a dedicated model, the framework i…
Background & analysis
On October 16, 2026, Tencent Hunyuan announced and open-sourced Hy Memory 2.0, an agent memory framework. According to IT之家, the release covers three layers of capability: a dedicated model, the memory framework itself, and the MemHub application. These three layers can be used independently or combined into a complete Memory solution.
The framework layer handles the storage, updating, and evolution of three types of memory: chat, work, and code. Developers can connect it to an Agent through an SDK. Tencent says the framework is the first in the industry to propose both work memory and code memory.
On evaluation, Tencent claims Hy Memory ranks first against open-source frameworks including Mem0, MemOS, EverOS, OpenViking, RepoMemory, and Simple RAG across 4 Benchmarks spanning chat, work, and code scenarios. For work memory, it scored an average of 55.31% on the travel planning and combined shopping tasks in MemoryArena, leading the runner-up EverOS by 10.2 percentage points. For code memory, the reported figure for code question-answering accuracy on CodeMemoryBench was cut off in the source material and could not be fully confirmed.
The report also notes that Hy Memory is already in use across multiple business lines inside Tencent, including finance, news, knowledge management, office Agent, social entertainment, and gaming. As of the report's publication, public information stops at the open-source release and the evaluation results above, with no further updates disclosed.
AI-generated from 1 reports · updated 9 hours ago
Latest turnTencent released Hy Memory 2.0 this week: an Apache-2.0 open framework, a 30B memory model, and MemHub, an app that shares one memory across AI tools. A developer maintaining a competing open-source memory library, who disclosed that conflict, benchmarked it on LongMemEval across two failure modes — stale facts and misattributed statements — and reports that 32 of Hy's 34 misses were "I do not have that information" answers.
Related AI tools
- Heat index
- 171
- All sources
- 2
- All reports
- 2
- First seen
- 10 hours ago
Reports on this story headlines open the original
Tencent released Hy Memory 2.0 this week: an Apache-2.0 open framework, a 30B memory model, and MemHub, an app that shares one memory across AI tools. A developer maintaining a competing open-source memory library, who disclosed that conflict, benchmarked it on LongMemEval across two failure modes — stale facts and misattributed statements — and reports that 32 of Hy's 34 misses were "I do not have that information" answers.
DEV Community · AIAI score 52
Tencent Hunyuan has announced and open-sourced Hy Memory 2.0, an agent memory framework released alongside a dedicated model and the MemHub application. Its framework layer handles storing, updating and evolving chat, work and code memory, and developers can plug it into agents through an SDK. Tencent says it tops Mem0, MemOS and other open-source frameworks across four benchmarks in those three scenarios.
IT之家AI score 64
Other stories people are talking about
- Heat index 1889RisingMeta open-sources Muse Spark Open and partners with Apple for local runs35 sources
- Heat index 848Anthropic IPO Valuation Talked Down to About $1.5 Trillion18 sources
- Heat index 763SurgeReflection's Beam Goes Open-Weight, Lands on Together AI and More10 sources
- Heat index 680RisingChina proposes adding 23 AI technologies to export restriction catalog9 sources
- Heat index 638RisingNvidia's Nemotron 3.5 Family Lands on Ollama and Baseten13 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion