wowsuch ai
Live

Tencent HunyuanModel

Hy-Memory-30B-A3B

Hy-Memory-30B-A3B is a compact long-term-memory writer model with approximately 3B active parameters, designed to decide what information to retain and how to update persistent memory. Its weights are available on Hugging Face, and the model can be deployed with vLLM and called through an OpenAI-compatible API.

What it does

Developers can use it to extract, write, and update user memories in frameworks such as Mem0, MemOS, and Hy-Memory.

Availability
Model weights, configuration, and model card on Hugging Face
Pricing model
Not established
Deployment
Deploy with vLLM; call through an OpenAI-compatible API
Privacy
Not established
Licence
Apache-2.0
Capability
extractiontext generation
Modality
text
Access
open weightsAPI
Deployment
on-premises
Pricing
unknown
Target user
developerresearcher

Evidence

9 claims
  • Hy-Memory-30B-A3B is open-sourced on Hugging Face. [1]
    * *<mark>[new]</mark>* 🔥🔥🔥 **[2026-10]**: 🚀 **Open-sourced** quantized **FP8** (**[Hy-Memory-30B-A3B-FP8](https://huggingface.co/tencent/Hy-Memory-30B-A3B-FP8)**) and **GGUF** (**[Hy-Memory-30B-A3B-GGUF](https://huggingface.co/tencent/Hy-Memory-30B-A3B-GGUF)**) versions of **Hy-Memory-30B-A3B** on Hugging Face!
    GitHub new repositories. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The model is trained specifically for high-precision long-term memory extraction and conflict reconciliation. [1]
    > **TL;DR:** Trains a 30B-A3B MoE model via attributed reinforcement learning specifically for high-precision long-term memory extraction and conflict reconciliation.
    GitHub new repositories. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • It is a 30B-A3B MoE model. [1]
    > **TL;DR:** Trains a 30B-A3B MoE model via attributed reinforcement learning specifically for high-precision long-term memory extraction and conflict reconciliation.
    GitHub new repositories. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • Its technical report describes attributed reinforcement learning for long-term memory writing. [1]
    * *<mark>[new]</mark>* 🔥🔥🔥 **[2026-10]**: 🚀 **Open-sourced** **[Hy-Memory-30B-A3B](https://huggingface.co/tencent/Hy-Memory-30B-A3B)** on Hugging Face along with its [technical report](https://arxiv.org/abs/2610.22196) on attributed reinforcement learning for long-term memory writing!
    GitHub new repositories. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The model is designed to execute the write path inside external memory frameworks. [2]
    **Hy-Memory-30B-A3B** is a compact long-term-memory writer model (~3B active parameters) developed by Tencent Hunyuan to execute the write path inside external memory frameworks.
    Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • It decides what information to retain, how to represent it, and how new evidence should update persistent memory. [2]
    It directly replaces the general-purpose LLM inside **Mem0**, **MemOS**, and **Hy-Memory** (covering Chat, Working, and Coding scenarios), deciding what information to retain, how precisely to represent it, and how new evidence should update the persistent memory bank.
    Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • Model weights, configuration, and model card are available on Hugging Face. [2]
    - **Model Weights, Configuration, and Model Card:** [https://huggingface.co/tencent/Hy-Memory-30B-A3B](https://huggingface.co/tencent/Hy-Memory-30B-A3B)
    Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • The model can be deployed with vLLM and called through an OpenAI-compatible API. [2]
    Deploy the model with vLLM (tested with v0.31.0) using `--served-model-name hy-memory`, then call the OpenAI-compatible API.
    Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights.
  • It is intended to replace the general-purpose LLM inside Mem0, MemOS, and Hy-Memory. [2]
    It directly replaces the general-purpose LLM inside **Mem0**, **MemOS**, and **Hy-Memory** (covering Chat, Working, and Coding scenarios), deciding what information to retain, how precisely to represent it, and how new evidence should update the persistent memory bank.
    Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights.

Sources

2
1
GitHub new repositoriesOwner-controlled
https://github.com/Tencent-Hunyuan/Hy-Memory Retrieved 2026-10-28 00:51 UTC
Verified
2
Hugging FaceOwner-controlled
https://huggingface.co/tencent/Hy-Memory-30B-A3B Retrieved 2026-10-28 00:51 UTC
Verified

Versions

15
VersionEffectiveChangeChanged fieldsState
152026-10-28 00:51 UTCmaterial changecapability, access, availability, deploymentCurrent
142026-10-26 01:59 UTCmaterial changecapability, access, availability, deploymentCurrent
132026-10-25 02:17 UTCmaterial changecapability, access, availability, deploymentCurrent
122026-10-24 03:11 UTCmaterial changecapability, access, availability, deploymentCurrent
112026-10-23 04:10 UTCmaterial changecapability, access, availability, deploymentCurrent
102026-10-22 05:01 UTCmaterial changecapability, access, availability, deploymentCurrent
92026-10-21 05:58 UTCmaterial changecapability, access, availability, deploymentCurrent
82026-10-20 06:57 UTCmaterial changepricingCurrent
72026-10-20 06:57 UTCmaterial changecapability, access, pricing, availability, deploymentCurrent
62026-10-19 07:47 UTCmaterial changecapability, availabilityCurrent
52026-10-19 07:47 UTCmaterial changecapability, access, pricing, availability, deploymentCurrent
42026-10-18 08:46 UTCmaterial changepricing, availabilityCurrent
32026-10-18 08:46 UTCmaterial changecapability, access, availability, deploymentCurrent
22026-10-17 09:45 UTCmaterial changecapability, access, availabilityCurrent
12026-10-16 10:38 UTCpublicationNoneCurrent

In briefs

2

Open in the catalog

Original catalog content is licensed under CC BY 4.0, attributed to wowsuchai. Source quotations and linked content keep their original rights.