} ?>
(Yicai) Sept. 8 -- Chinese internet titan Tencent Holdings debuted its first large language artificial intelligence model Hunyuan at its Global Digital Ecosystem Summit in Shenzhen.
Hunyuan is open to enterprises in China for testing and building apps via application programming interfaces on Tencent Cloud, the Shenzhen-based company announced yesterday.
Hunyuan has over 100 billion parameters, with more than two trillion tokens in pre-training data, Tencent noted. It supports a wide array of functions, including the creation of images, copywriting, text recognition, and customer service, and will be applied in key industries, such as finance, public services, social media, e-commerce, transportation, and games.
Hunyuan has been connected to 50 of Tencent’s products, including Tencent Cloud, Tencent Marketing Solutions, Tencent Games, Tencent fintech services, Tencent Meeting, Tencent Docs, Weixin Search, QQ Browser, and other core offerings.
Tencent’s Vice President Jiang Jie said at the Global Digital Ecosystem Summit that Hunyuan outperforms ChatGPT-3.5 in many respects, but still falls behind ChatGPT-4. Hunyuan can write up to 4,000-word articles, while ChatGPT-3.5 can reach only 1,000 words, Jiang noted.
Large Chinese tech firms must develop LLMs in a gradual way to make technical breakthroughs step by step and be frank with the gap with foreign tech companies, Jiang pointed out.
On June 19, Tencent launched a store for corporate and industrial LLMs to offer one-stop model-as-a-service services to build LLMs and intelligent applications, Dowson Tong, senior executive VP of Tencent and chief executive of Tencent Cloud and Smart Industries Group, said a few months ago.
It is probably a better choice for companies to build their exclusive LLMs based on an industry-specific LLM rather than using a general-purpose product, Tong added, noting that exclusive LLMs have fewer parameters than general-purpose LLMs, cost less in terms of training and reasoning, and are easier to optimize.
Editor: Futura Costaglione