Hunyuan
Tencent's open AI models for video, images and 3D
- Made by
- Tencent
- Category
- Model
- Last change
- Nov 2, 2026
Changelog
November 2026
- Launch
Tencent Hunyuan introduced Hyra 2.0, a research agent that builds reusable tools, models and workflows to speed up future research.
What this means for you: Tencent's Hunyuan team introduced Hyra 2.0, a research agent (an AI that carries out tasks on its own). It builds reusable tools, models and workflows so future research goes faster.
How to use it: Tencent's write-up is at hy.tencent.com/research/hyra2.
October 2026
- New model
Hunyuan open sourced Hy-SimulMT, a real-time Chinese and English simultaneous interpreting model with voice cloning and about 3 seconds of delay.
What this means for you: This is a model that translates spoken Chinese and English into the other language as the person talks, with about 3 seconds of delay. It can also clone the speaker's voice in real time.
Why it matters: It follows the Sep 24 launch of Tencent Hy Translation, an app with voice translation in 33 languages.
How to use it: Download it from Hugging Face, ModelScope or GitHub under the name Hy-SimulMT. It is open source.
- Launch
Hunyuan released Hy-Nomas, an open source interactive video world model with text and camera control, runnable on consumer GPUs.
What this means for you: Hy-Nomas is a free, open source AI model that makes interactive video worlds. You steer it with text prompts and camera paths, and it keeps a memory of the world during long clips. It is built to run on consumer graphics cards, and a lightweight add-on method (LoRA) adapts it to specialized fields such as robotics.
How to use it: Try the demo at hy-nomas.hy.tencent.com/demo. The code is on GitHub at Tencent-Hunyuan/Hy-Nomas, and the model is on Hugging Face at tencent/Hy-Nomas.
- Launch
Hunyuan released Hy Memory 2.0, an open-source memory model and framework, plus MemHub, a CLI and Mac app that syncs memory across AI tools.
What this means for you: Hy Memory 2.0 is a free, open-source tool that keeps track of your chats, work and code so an AI assistant can remember them. MemHub is a command-line tool (typed commands) and a Mac app that syncs that memory across more than 10 AI tools.
How to use it: The SDK and MemHub are on GitHub at github.com/Tencent-Hunyuan/Hy-Memory. The model, Hy-Memory-30B-A3B, is on Hugging Face at huggingface.co/tencent/Hy-Memory-30B-A3B. MemHub runs on Mac.
I use several AI tools and just learned Tencent Hunyuan's Hy Memory 2.0 and its MemHub Mac app can sync memory across 10+ AI tools. Help me work out what I would want an AI to remember about my chats, work and code, and how to organize it. My tools are: [list tools].
September 2026
- Launch
Hunyuan launched Tencent Hy Translation, an app with voice, photo and offline on-device translation in 33 languages, live in 12 countries and regions.
- New model
Hunyuan released Hy Image3.5 preview, with text-to-image and image editing up to 2K, priced at $0.024 per image on Tencent Cloud API.
- New model
Hunyuan released AuK, an open-source model for speech generation and editing, plus AuK-Flash, which runs about 4.5 times faster.
- New feature
Hunyuan updated Hy4 preview to reduce long thinking and over-verification on complex tasks, with fewer turns and lower token use, live for everyone.
August 2026
- New model
Hunyuan released Hy4 preview, an open source model with 770B parameters, 49B active and a 1M context window.
Related tools
- DeepSeekChinese lab's open-weight models and free chat app
- OllamaDownload and run open AI models like Llama, Gemma and Qwen on your own computer
- OpenRouterOne account and one API to use hundreds of AI models from every major lab
- ReplicateRun open AI models in the cloud with one click or one line of code
- KimiMoonshot's open-weight models and chat app
- LM StudioA desktop app to find, download and chat with open AI models offline
