Organisation
Tencent
The Hunyuan family of language, image and video models, from the company behind WeChat.
Tencent, the Chinese gaming and social-media group behind WeChat, develops the Hunyuan family of models, which it first unveiled in 2023 as an enterprise-focused, cloud-delivered system rather than a consumer chatbot. That enterprise-first approach reflected its position as a cloud and software provider, and Hunyuan grew into a series spanning language, image and video generation. By 2026 it was releasing open-weight mixture-of-experts models under permissive licences, competing with rivals such as DeepSeek and Zhipu on token efficiency and licensing rather than on raw parameter count. Its consumer assistant, Yuanbao, has trailed ByteDance's Doubao and other rivals in China's crowded assistant market.
- Category
- Chinese AI labs
- Founded
- 1998
- HQ
- Shenzhen, CN
- Key people
- Ma Huateng
Appears alongside
Featured in threads
Tracks
- Models & capabilities 7
- Open weights & ecosystem 6
- Security & misuse 1
- Compute & infrastructure 1
- Government & policy 1
- Money & business 1
- Benchmarks & progress 1
Researchers build an AI-assisted WeChat worm
The demonstration, called WeWorm, spread via missed WeChat calls on iOS and Android without user interaction; Tencent patched the underlying flaw before disclosure and no exploitation in the wild was reported.
Security & misuse
Tencent open-sources its Hy4-preview language model
The 770-billion-parameter model, with 49 billion active, is free on Tencent's coding tools for two weeks and priced under a dollar per million input tokens after that.
Models & capabilities · Open weights & ecosystem
First Nvidia H200 shipments reach China
ByteDance and Tencent each received about 10,000 chips, roughly 13% of the ceiling their US licences allowed, with Beijing reportedly requiring most of the rest to stay in Hong Kong.
Compute & infrastructure · Government & policy
Lovable raises $400M at a $13.3bn valuation
The Stockholm vibe-coding startup's valuation roughly doubled in eight months; TechCrunch reported its annualised revenue was approaching $600m.
Money & business
Tencent releases Hy3, a 295B open-weight MoE model
The 295B-parameter, 21B-active model was released under Apache 2.0 and, Tencent said, matched much larger rivals GLM-5.2 (753B) and DeepSeek-V4-Pro (1.6T) while using far fewer tokens.
Open weights & ecosystem · Models & capabilities · Benchmarks & progress
Tencent releases Hunyuan 2.0
A 406-billion-parameter mixture-of-experts model with 32 billion active parameters and a 256,000-token context window, released in Think and Instruct variants.
Models & capabilities · Open weights & ecosystem
Tencent open-sources Hunyuan Image 3.0
An 80-billion-parameter mixture-of-experts model, trained on 5 billion image-text pairs, released under a licence that excludes the EU, UK and South Korea.
Open weights & ecosystem · Models & capabilities
Tencent releases HunyuanVideo
At 13 billion parameters, Tencent called it the largest open-weight video model, and said blind evaluators rated its motion quality above Runway Gen-3 and Luma 1.6.
Open weights & ecosystem · Models & capabilities
Tencent open-sources Hunyuan-Large MoE model
Tencent said the 389B-parameter, 52B-active MoE model beat Llama 3.1 405B on MMLU and MATH despite far fewer active parameters, and released a technical report alongside the weights.
Open weights & ecosystem · Models & capabilities
Tencent releases first Hunyuan large model
Unveiled at Tencent's Global Digital Ecosystem Summit, the mixture-of-experts model exceeded 100 billion parameters and launched for enterprise access via Tencent Cloud, not consumers.
Models & capabilities
In the commentary
Pieces from around the web that discuss Tencent. External links.
Also mentioned in 4 entries
Referenced in passing — Tencent isn't the main subject of these.