Today for AI
HOT RADAR
agentHEAT 10.9°

AI CURATED HOT STORY · 10/16/2026

Tencent Open-Sources Hy Memory 2.0 to Enhance Agent Working Memory

1 reports archived1 independent sourcesupdated 10/16/2026, 18:18:46
Synthesis & Latest Updates
1 Sources Cross-Validated

Tencent has open-sourced Hy Memory 2.0, an agent memory framework introducing dedicated 'working' and 'code' memory modules to maintain context in long-horizon tasks. The system outperforms competitors like Mem0 on multiple benchmarks and reduces token usage by nearly 40% in chat scenarios. The SDK, MemHub, and specialized model weights are now available on GitHub and Hugging Face.

LATEST/Tencent has open-sourced Hy Memory 2.0, an agent memory framework introducing dedicated 'working' and 'code' memory modules to maintain context in long-horizon tasks. The system outperforms competitors like Mem0 on multiple benchmarks and reduces token usage by nearly 40% in chat scenarios. The SDK, MemHub, and specialized model weights are now available on GitHub and Hugging Face.

HEAT TRENDHourly heat curve

4 fully observed hours
Now
10.9
Peak
11.819:00
24h change
–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

TIMELINECoverage timeline

Total 1 reports · Latest first
  1. IT之家 · 智能时代T2·78 pts
    • First open-source framework to propose separated 'working memory' and 'code memory' architectures, enabling cross-time information integration.
    • Ranked first across four benchmarks including MemoryArena and CodeMemoryBench, leading the runner-up by over 10 percentage points.
    • Reduced per-question token usage from 1,159 to 703 (a 39.3% cost drop) in chat scenarios via event-contextualization techniques.
Tencent Open-Sources Hy Memory 2.0 to Enhance Agent Working Memory | Today for AI