Tencent Hunyuan upgrades and open-sources Hy Memory 2.0

AT A GLANCE

The report states that Tencent Hunyuan open-sourced Hy Memory 2.0 added working and code memory, and supports local memory sharing across AI tools through MemHub.

Article

Tencent Hy Memory 2.0 open-sourced: 10 tools share memory, with even rolled-back code traceable AI Beating AI breaking news, Tencent Hunyuan has upgraded Hy Memory to 2.0, open-sourcing the memory framework, dedicated model, and MemHub application all at once.

This year,May launched 1.0, which primarily handles chat memory. It can track changes to the same matter over time, identify and resolve contradictions, and infer user habits that were not explicitly stated from multiple conversations. When users change their preferences, the system can retain old records while remembering the latest decision. 2.0 adds working memory and code memory. It records how far a task has progressed, which approaches have been used, and what the final result was. When it encounters a similar task, it also checks whether past experience applies. Code memory goes a step further by combining with Git records to track the modification history of files and functions. Even if a piece of code is later rolled back, it can retrieve the changes made at the time and identify which approaches had previously been tried.

The accompanying MemHub connects memory across different AI tools. Cursor, Codex, Claude Code, and other 10 tools can share a single local memory. After the user grants authorization, MemHub can import historical conversations and continuously synchronize new records. For example, after project specifications are established in Cursor, switching to Codex to continue development can retrieve them directly without needing to explain them again.

Tencent also trained a dedicated model,Hy-Memory-30B-A3B, with a total of 30000000000 parameters, of which only about 3000000000 are activated per inference. It is specifically responsible for filtering, organizing, and updating memory, and can also integrate with existing frameworks such as Mem0 and MemOS. Tencent tested four memory systems, with an average score of 55.49, slightly higher than Opus 5’s 54.62.

Original link https://m.theblockbeats.info/flash/372245

Key points

Dongcha Beating AI news flash reports that Tencent Hunyuan upgraded Hy Memory to 2.0 and open-sourced the memory framework, dedicated model, and MemHub application.
The report states that Hy Memory 1.0 primarily handles conversational memory and can track changes, handle contradictions, and infer user habits.
The report states that 2.0 adds working memory and code memory, and can record task progress, approaches used, and results.
Code memory can combine Git records to track the modification history of files and functions and find code changes that were later rolled back.
MemHub enables Cursor, Codex, Claude Code, and other 10 tools to share one local memory; after user authorization, it can import historical conversations and synchronize new records.
The dedicated model Hy-Memory-30B-A3B has a total of 30000000000 parameters, with only about 3000000000 activated per use, and is responsible for filtering, organizing, and updating memories.
The report states that the model can integrate with existing frameworks such as Mem0 and MemOS.
The report claims that Tencent tested four memory systems, with an average score of 55.49 slightly higher than Opus 5 at 54.62。

AI-assisted interpretation

The following is analysis, separate from reported facts. Verify important claims independently.

This system attempts to enable AI tools to remember conversational preferences, task progress, and code modification history. Information established by a developer in one tool can continue to be used in other supported tools through MemHub; code attempts rolled back by Git may also be traceable.

Why it matters to readers

Cross-tool memory sharing can reduce the need for users to repeatedly explain project context, conventions, and previous attempts; code memory may also help developers understand earlier modification and rollback processes.

If you alternate among multiple AI coding tools, the shared-memory capability described for MemHub may allow tools to carry forward previous context, but authorization, privacy, and the actual scope of compatibility should still be confirmed before use.

Risks and unknowns

The evidence is a media news flash and does not include a Tencent official announcement, code repository, or model-weight address; therefore, it cannot independently confirm whether the open-sourced content is accessible.
The original text did not list all 10 supported tools, nor did it specify the precise scope of compatibility.
The original text did not explain the privacy and security mechanisms for importing historical conversations, local storage, authorization revocation, or synchronization.
55.49 and 54.62’s testing methods, datasets, evaluation criteria, and independent reproducibility were not explained.

Tencent Hunyuan upgrades and open-sources Hy Memory 2.0

Article

AI-assisted interpretation

Why it matters to readers

Risks and unknowns

Related Developments

Related concepts

Keep following

Security Center