Where is Tencent (Hunyuan AI Model Family) headquartered?

Tencent (Hunyuan AI Model Family) is headquartered in Shenzhen, China. The company has a team of 500+ people.

What does Tencent (Hunyuan AI Model Family) do?

Tencent (Hunyuan AI Model Family) operates in the Foundation Models / LLMs sector. Tencent's Hunyuan (混元) is the company's proprietary, full-stack self-developed foundation model family spanning LLM, image, video and 3D generation, with the Hy3 model released July 2026 and a larger Hy4 model planned. Tencent Hunyuan provides enterprise-grade, multimodal AI models (text, image, video, 3D) that are deeply integrated across Tencent's massive product ecosystem (WeChat, Yuanbao, WorkBuddy, CodeBuddy, ima, QQ, Tencent Docs) and offered via Tencent Cloud, combining open-source Apache-2.0 weights with cost-efficient MoE architecture.

When was Tencent (Hunyuan AI Model Family) founded?

Tencent (Hunyuan AI Model Family) was founded in 1998. It is based in Shenzhen, China.

Who is the CEO of Tencent (Hunyuan AI Model Family)?

Pony Ma (Ma Huateng / 马化腾) is the CEO of Tencent (Hunyuan AI Model Family). The company was founded in 1998.

Where is Tencent (Hunyuan AI Model Family) located?

Tencent (Hunyuan AI Model Family) is headquartered in Shenzhen, China.

What funding stage is Tencent (Hunyuan AI Model Family) at?

Tencent (Hunyuan AI Model Family) is currently at the Public stage.

How many employees does Tencent (Hunyuan AI Model Family) have?

Tencent (Hunyuan AI Model Family) has approximately 500+ employees.

Tencent (Hunyuan AI Model Family)

Category: Foundation Models / LLMs

Tencent's Hunyuan (混元) is the company's proprietary, full-stack self-developed foundation model family spanning LLM, image, video and 3D generation, with the Hy3 model released July 2026 and a larger Hy4 model planned. Tencent (Hunyuan AI Model Family) was founded in 1998. The company is led by Pony Ma (Ma Huateng / 马化腾). Based in Shenzhen, China. Team size: 500+. Latest round: IPO. Key investors include Publicly traded on HKEX (0700), Prosus/Naspers are major shareholders (~22.8% as of April 2026, down from historical 46.5%).

Founded: 1998
Headquarters: Shenzhen, China
Team size: 500+

Value proposition

Tencent Hunyuan provides enterprise-grade, multimodal AI models (text, image, video, 3D) that are deeply integrated across Tencent's massive product ecosystem (WeChat, Yuanbao, WorkBuddy, CodeBuddy, ima, QQ, Tencent Docs) and offered via Tencent Cloud, combining open-source Apache-2.0 weights with cost-efficient MoE architecture.

Products and solutions

Hunyuan Hy3 (295B MoE, 21B active, 256K context, Apache 2.0, released July 6 2026), Hy3 Preview (April 2026), planned larger Hy4, HunyuanImage 3.0, HunyuanVideo, Hunyuan 3D, HunyuanOCR, Hy3-FP8 quantized variant.

Unique value

Hy3 combines a 295B-parameter Mixture-of-Experts architecture with 21B active parameters and 256K context window, offering hybrid fast/slow reasoning modes (no_think, think_low, think_high) for cost-efficient agentic performance, released open-source under Apache 2.0.

Target customer

Enterprises via Tencent Cloud, developers (open-source community on Hugging Face/ModelScope/GitCode), and Tencent's own consumer products (WeChat, Yuanbao, QQ).

Industries served

Enterprise productivity (WorkBuddy), software development (CodeBuddy), consumer AI assistant (Yuanbao), knowledge management (ima), cloud computing (Tencent Cloud), content generation (image/video/3D).

Technology advantage

Full-stack self-developed model family with MoE architecture; open-source Apache-2.0 weights; hybrid reasoning modes; deep integration with Tencent's WeChat-scale distribution; NVIDIA TensorRT-LLM optimization and H20 GPU training; ranked #1 on OpenRouter global LLM usage leaderboard within a week of Hy3 launch (68x API call surge).

How they differentiate

Unlike pure-play model startups, Tencent differentiates through WeChat-scale consumer distribution and deep product integration (Yuanbao, WorkBuddy, CodeBuddy, ima, QQ), combined with open-source Apache-2.0 weights and cost-efficient MoE design that topped OpenRouter's global usage leaderboard.

Main competitors

DeepSeek, Alibaba Qwen, Baidu Ernie, ByteDance Doubao, Moonshot Kimi, Zhipu AI GLM

Key partnerships

NVIDIA (TensorRT-LLM inference optimization, H20 GPU training), Hugging Face, ModelScope, GitCode (open-source distribution), OpenRouter (API availability).

Notable customers

Tencent's own products (WeChat, Yuanbao, WorkBuddy, CodeBuddy, ima, QQ, QQ Browser, Tencent Docs, Tencent LearnShare) plus enterprise customers via Tencent Cloud.

Major milestones

2023: Hunyuan LLM launched (Hy1), Dec 2025: Tencent restructured AI ops, appointed Shunyu Yao as Chief AI Scientist, April 2026: Hy3 Preview released, July 6 2026: Hy3 officially released (295B MoE, Apache 2.0), 68x API surge, #1 on OpenRouter, Aug 2026: Tencent announced plans for larger Hy4 model.

Growth metrics

Hy3 weekly usage jumped 68-fold vs predecessor after moving from preview to formal version; ranked #1 on OpenRouter global LLM usage leaderboard within one week; WorkBuddy recorded 20M+ PC visits in June 2026; Tencent Q2 2026 revenue reached 204.79 billion yuan.

Market positioning

One of China's top-tier foundation model players, competing with DeepSeek, Alibaba Qwen and Baidu Ernie; leads in consumer distribution via WeChat ecosystem while offering open-source enterprise models through Tencent Cloud.

Geographic focus

China (primary), with expanding global/overseas rollout of Hy3 via WorkBuddy, Tencent Cloud and developer platforms.

Patents and IP

Open-sourced under Apache 2.0 on Hugging Face, ModelScope, GitCode and CNB; Hy3 and Hy3-FP8 weights publicly released.

About Pony Ma (Ma Huateng / 马化腾)

Co-founder, Chairman & CEO of Tencent Holdings; founded Tencent in 1998. Built WeChat and China's largest internet ecosystem.

Latest news about Tencent (Hunyuan AI Model Family)

Tencent Plans Larger Hy4 Model Following Hy3 Weekly Usage Surge of 68x

More Foundation Models / LLMs companies

Official website: https://hunyuan.tencent.com/