When you buy through affiliate links in our content, we may earn a commission at no extra cost to you. Learn 
As the ChatGPT hype and traffic begin to die down, the generative AI market is becoming increasingly more diverse. Earlier this month, Chinese tech giant Tencent announced the release of Hunyuan, a multimodal large language model (LLM) available through the Tencent public cloud platform.
Hunyuan supports over 100 billion parameters and was trained with over 2 trillion tokens, which are designed to generate natural language in Chinese.
Senior Executive Vice President of Tencent and CEO of Tencent Clouting and Smart Industries Group (CSIG), Dowson Tong, said in the announcement blog post:
“Enterprises in China may now access Hunyuan via Tencent’s public cloud platform and fine-tune it to their specific needs. The platform features strong Chinese language processing abilities, advanced logical reasoning, and comes with reliable task execution abilities.”
It also supports a range of use cases, including image creation, copywriting, text recognition, and customer service.
At a high level, the release illustrates that China is entering into the generative AI development race, coming just weeks after Baidu and SenseTime Group announced the launch of their own ChatGPT-style chatbots.
How Does Tencent Stack Up Against ChatGPT?
Today, ChatGPT remains the most popular chatbot in use, even if it has shown signs of losing users. However, the tool hasn’t been able to penetrate the Chinese market after Beijing banned the virtual assistant earlier this year for containing content that was illegal in the country.
This ban has left the market largely undeserved, particularly when considering that researchers have criticized the capabilities of ChatGPT in languages other than English.
Considering that 1.4 billion people speak Chinese, Hunyuan has the potential to be one of the biggest LLM if it lives up to expectations.
For this reason, the launch of an LLM that’s designed specifically to support natural language processing (NLP) and generation in Chinese is a game changer in the market. Tencent claims that its model can converse in English and Chinese and is “better” than ChatGPT at generating long text outputs and solving particular math problems.
However, it’s not just the emphasis on Chinese language output that differentiates Hunyuan – it also offers a high standard of performance. Its 100 billion parameters exceed that of Meta Llama 2‘s 70 billion. While it’s significantly behind GPT-3.5’s 175 billion parameters, it still has the bandwidth to be competitive.
It’s worth noting that Tencent also claims that Hunyuan experiences 30% fewer hallucinations than Llama 2.
Likewise, its ability to generate images provides an edge over ChatGPT, which, as of writing, cannot.
Generative AI for the Enterprise
Another interesting point of differentiation from Hunyuan has been Tencent’s enterprise focus. Back in June, the organization released Tencent Cloud MaaS, a Model-as-a-Service (MaaS) solution designed to provide organizations with models that were customized to their industry.
This includes custom models for sectors including finance, healthcare, media, and education. Hunyuan can integrate with these tools and a wide array of tools across the Tencent product ecosystem, including:
- Tencent Cloud
- Tencent Marketing Solutions
- Tencent Games
- Tencent finch systems
- Tencent Meeting
- Tencent Docs
- Weixin Search
- QQ Browser
Tencent’s approach toward helping enterprises streamline the development of industry-specific AI models based on a mix of proprietary and open-source models is a key point of differentiation from OpenAI.
Currently, OpenAI hasn’t developed AI models catered to specific industries, but a consumer-grade chatbot in ChatGPT and a more generalist enterprise solution in the form of ChatGPT Enterprise.
Is AI Development Decentralizing?
The release of Hunyuan also highlights that the generative AI market is gradually decentralizing. While development and research on AI-driven solutions have taken place worldwide for years, the early days of LLM development were largely dominated by a handful of vendors in Silicon Valley.
For instance, OpenAI, Anthropic, Meta, and Google, which have contributed to leading chatbots like ChatGPT, DALL-E 2, Bard, Llama 2, and PaLM 2, are all based in California.
The release of Hunyuan, following the launch of chatbots from Baidu and SensetimeGroup, highlights that China is a key entity to watch in terms of AI development, particularly for organizations catering to the Mandarin-speaking market.
Although vendors like Google, OpenAI, and Meta will remain key players for years to come, the release of Hunyuan and the UAE’s Falcon 180-B highlight that the development of language models is becoming increasingly decentralized.
Building the Chinese Generative AI Market
Ultimately, Hunyuan’s launch reaffirms that there is a viable generative AI arms race occurring between the U.S. and Chinese markets that will be decided based on specialization.
The way to compete against entrenched tools like ChatGPT is to create solutions that are better tailored toward certain use cases and industries.
