Tencent Holdings, the prominent Chinese tech company, has officially introduced its expansive language AI model known as 'Hunyuan' for commercial applications. This announcement occurred at a conference in Shenzhen, where Tencent demonstrated how Hunyuan is integrated into more than 50 of its products and services.

Hunyuan boasts an impressive 100 billion parameters and was trained on a dataset containing over 2 trillion tokens, giving it significant computational power and positioning it as one of the most robust language models in China.

Tencent claims that Hunyuan outperforms OpenAI's ChatGPT in several aspects, including the generation of lengthy texts with thousands of words and solving complex mathematical problems. The model is versatile, and capable of engaging in conversations in both Chinese and English, making it suitable for a wide range of applications, including image creation, copywriting, text recognition, and customer service.

Additionally, Tencent asserts that Hunyuan exhibits 30% less 'hallucination' compared to Meta Platform Inc's Llama 2 model. 'Hallucination' refers to the generation of incorrect information presented as factual by AI models, suggesting that Hunyuan demonstrates a higher level of accuracy and reliability.

It's important to note that there has been no independent verification of Tencent's claims, and neither OpenAI's ChatGPT nor Meta's Llama 2 has issued immediate responses to Tencent's assertions.

Why does it matter?

The introduction of Hunyuan reflects the broader trend in China, where several tech firms, including Baidu Inc and SenseTime Group, have recently unveiled their own AI models. Current data suggests that there are over 130 large language models in China, illustrating the intense competition within the country's AI sector. This underscores China's commitment, both in the private and governmental sectors, to establish itself as a leader in AI technology.