简体

About Us

Kunlun Tech has been deeply engaged in overseas market for more than 10 years, business coverage includes information distribution, social, entertainment, meta-universe, games and AIGC and other fields, including AGI and AIGC, overseas information distribution and meta-universe, investment and other three business segments. Markets throughout China, Southeast Asia, Africa, the Middle East, North America, South America, Europe and other places. Up to now, the global average monthly active users are nearly 400 million, and overseas revenue accounts for 92%.
As a leading artificial intelligence technology company in China, Kunlun Tech has built a diverse AI business matrix consisting of AI intelligent assistants, AI video, AI music and audio, AI games, AI social networking, and other related businesses.

Memorabilia

2008
2012
2015
2018
2019
2020
2021
2022
2023

Kunlun is founded

Kunlun starts with web game development and publishing games globally, and successively releases games such as "Kunlun World", "Girl’s Dream Factory", "Tale of Wuxia" among others. Kunlun is part of the first batch of Chinese developers to enter the overseas game market.

Kunlun gradually starts to develop self-developed mobile games

Relying on development and operating experience from web games, Kunlun gradually expands the business to envelop online mobile games and client games. In 2012, the mobile game business is officially established, and Kunlun focuses on developing mobile games in-house starting in 2015.

Kunlun lists in A-share, develops an investing plan

Kunlun establishes in the ChiNext stock market on January 21, 2015. Based on the game business, it strategically invests in Opera, Grindr and Xianlaihuyu, in order to widen the focus and to capture opportunities in the fields of Fintech, live broadcast, e-commerce, and two dimensions.

Kunlun keeps expanding business scope

Kunlun is committed to becoming the world's leading social network company, has listed the recently acquired Opera on the NASDAQ in New York, and the new independent brand GameArk is officially launched, to further expand its overall business scope.

Kunlun aims to be a global Internet Platform Company

Several diversified business segments have been initially formed, including the information distribution platform Opera, the global game development-and-operation integrated business GameArk, and the leisure and entertainment platform Xianlaihuyu, so as to transform into a global internet platform company.

Kunlun reinforces the positioning on Internet Platform Company

Through acquisitions, Kunlun now maintains a 53.88% stake in Opera and becomes the controlling party of the global internet traffic giant, further clarifying and strengthening the strategic positioning of the global internet platform.

Kunlun enhances its global influence thoroughly

Kunlun acquires Star Group to expand into audio-based social networking; Opera and Star Group are merged into Kunlun financial statements to further enhance the global internet platform strategy and thoroughly increase its global reach and influence.

ALL in AIGC

In 2022, Kunlun Tech will start research and development in ChatGPT, image, video and other AIGC technologies; On December 15, 2022, five projects including the Chinese class GPT-3 model, the conversational chatbot based on the class GPT-3, SkyMusic and SkyText were opened.

All in AGI and AIGC

In April 2023, led by KunlunTech research and development, China's first truly intelligent emerging domestic large language model "Tiangong" 3.5 opened the private test.

Open-Source Model Series

Skywork/SkyReels-A1

Given an input video sequence and a reference portrait image, we extract facial expression perception feature points from the video and use them as motion descriptors to transfer the expressions onto the portrait image. Leveraging a conditional video generation framework based on DiT (Diffusion Transformer), these facial expression perception feature points are directly integrated into the input latent space.

Skywork/Skywork-VL-Reward-7B

We achieved a state-of-the-art (SOTA) score of 73.1 on VL-RewardBench and a high score of 90.1 on RewardBench. Furthermore, the Maximum a Posteriori Policy Optimization (MPO) we trained on Skywork-R1V-2.0 has further verified the effectiveness of the model.

Skywork/Skywork-R1V2-38B

Skywork-R1V2-38B is a state-of-the-art open-source multimodal reasoning model that has achieved top-tier performance across multiple benchmark tests: On MMMU, it scored 73.6%, the highest among all open-source models to date；On OlympiadBench, it achieved a score of 62.6%, outperforming other publicly available models by a significant margin.

Skywork/Skywork-SWE-32B

Skywork-SWE-32B achieved an accuracy of 38.0% in terms of pass@1 on the SWE-bench Verified benchmark, outperforming the previous open-source State-of-the-Art (SOTA) Large Language Model (LLM) — Qwen2.5-Coder-32B — which was built based on the OpenHands agent framework.

Skywork/Skywork-R1V3-38B

Skywork-R1V3-38B is the latest and most powerful open-source multimodal reasoning model in the Skywork-R1V series. Built on the foundation of InternVL-38B, it has achieved a significant breakthrough in pushing the boundaries of multimodal and interdisciplinary intelligence. Primarily leveraging post-training reinforcement learning algorithms to enhance its reasoning capabilities, R1V3 has attained state-of-the-art (SOTA) performance among open-source models across numerous multimodal reasoning benchmarks.

Skywork UniPic 1.0

Skywork-UniPic is a unified autoregressive multimodal model with 1.5 billion parameters, capable of handling three key vision-language tasks within a single architecture: image understanding, text-to-image generation, and image editing.

Matrix-Game 2.0

Matrix-Game-2.0 (1.8B) is an interactive world model that enables instant generation of long videos through a few steps of autoregressive diffusion.

Skywork/Matrix-3D

Matrix-3D leverages panoramic representation to generate a 3D world with wide coverage that is explorable in all directions, integrating conditional video generation and panoramic 3D reconstruction.

Skywork/Skywork-o1-Open-PRM-Qwen-2.5-1.5B

Unlike the mere replication of OpenAI's o1 model, the Skywork o1 Open model series not only demonstrates inherent capabilities in thinking, planning, and reflection in its outputs but also shows a significant improvement in reasoning ability in standard benchmark tests. This series represents a strategic advancement in artificial intelligence capabilities, elevating previously less robust foundation models to the state-of-the-art (SOTA) level in reasoning tasks.

More>>News

Kunlun Tech released the " SkyAgents" platform, zero code to build AI agents

2023.12.01

More details.

Kunlun Tech releases open source 13B high-quality commer...

2023.10.30

More details.

The Kunlun Tech MMORPG "Tower of the Holy Land" will ...

2023.09.07

More details.

Contact us

PR:
PR@kunlun-inc.com

IR:
ir@kunlun-inc.com

Report Offences:
tousu@kunlun.com

 

 
Add:Block B, Mingyang International Center, Beijing

Tel: 010-65210329

Fax: 010-65210399

昆仑万维科技股份有限公司 京ICP备09007496号-13 京公网安备11010102000432号
京ICP证080262号（署）网出证（京）字第205号